This achievement described by the DOE as “the rebirth of America’s nuclear industry” is part of a broader effort by the Trump administration to achieve its Nuclear Energy Dominance agenda.

Energy Secretary Chris Wright used the following words on April 15th of this year to describe this milestone.

“We can have reactors critical, meaning running their full nuclear systems. They’re not turning it into electricity, but that’s the simple part… We will have multiple small modular reactors running… in the next three months.”

Wright’s enthusiasm is contagious but his framing of criticality needs some interrogation as it inverts where the real engineering difficulty lies.

Subscribe now

Achieving criticality is an important milestone, but the engineering involved is comparatively straightforward. We have understood how to assemble fissile material into a self sustaining chain reaction for more than 80 years.

In just over two weeks in 1942, Italian-American physicist Enrico Fermi and his team arranged 42,000 bricks of ultra pure graphite and 40 tons of natural uranium into a spherical “pile” propped up by a massive wooden timber framework in an underground squash court at the University of Chicago. This was the world’s first man-made criticality.

The Antares test reactor used significantly more sophisticated materials: The fuel, HALEU containing TRISO fuel elements manufactured by BWXT, was arranged within a graphite moderated test core and if true to Antares conceptual design was regulated by control drums. Make no mistake though this was a barebones reactor.

The sodium heat pipes and heat exchanger that Antares intends to use to remove heat from the core and the closed nitrogen Brayton cycle turbine to convert that heat into electricity were absent from the experiment. The framing of this flash in the pan as having “reactors running” is a stretch.

So what’s the best way to understand the significance of achieving criticality in such a pared down system?

The combustor of a jet engine offers an illustrative analogy. Energy release is real but the combustor in isolation does not move an aircraft.

Without the precisely machined compressor feeding it air, the single crystal turbine blades extracting work from its exhaust alongside the multitude of systems and controls coordinating fuel delivery, lubrication, cooling and thrust management the combustor is simply a sophisticated flame thrower.

To run with the reliability required for commercial aviation these systems require years if not decades of design iteration, prototyping and testing.

A commercial reactor is fundamentally a heat engine, and the challenge lies in controlling, moving and converting significant quantities of heat into electricity.

Every component in that chain: fuel, cladding, pumps, valves, turbines, generators, control systems, instrumentation and safety systems, must function together with extraordinary precision under high temperatures, radiation fields, thermal cycling, and demanding operating conditions for years on end.

The June 4th criticality was an essential milestone to gather data validating reactor physics, fuel performance, and control system behaviour in the Antares core but it was a baby step on the long tedious march ahead towards a power producing reactor, let alone an economic one.

Nothing about this event has the ring of “dominance” about it. In the context of the Chinese approving 10GW of new large nuclear reactors per year and being responsible for 50% all new nuclear reactors currently under construction worldwide it was a footnote.

An empty vessel does however make the most noise.

The fact that a VC funded start-up got to this stage so quickly is certainly a regulatory breakthrough, but much less a technological one.

The company’s challenge now is traversing the valley of death that has claimed many a reactor venture before. It must prove that it can engineer the remainder of the reactor’s systems, satisfy regulators, operate its full systems on a Project Janus military base deployment, and finally attract commercial customers.

The latter may be the most challenging of all given the extreme diseconomies of scale of microreactors, but that analysis is for another time.

Like this post to keep us going critical

Oklo, which merged with Sam Altman’s SPAC in 2024 at an $850 million valuation, reached a peak market cap above $30 billion in 2025, largely on AI electricity demand expectations. Microsoft is reportedly paying $115/MWh for the revived Three Mile Island’s output under a 20-year contract, roughly double historic PJM wholesale prices.

The nuclear sector should not bite the hand that feeds it, but it should examine its palm lines very carefully, because there are now serious reasons to believe the AI demand narrative rests on questionable economics and over investment in an LLM-centric approach. While the technology has seduced many and has genuinely important use cases, limitations are emerging that may be intrinsic to LLM architecture itself.

The exploration of the evidence indicating an AI bubble is inspired by the accompanying podcast interview with returning guest David Helmer founder and managing partner at NTTW consulting. Stick around for the conclusion where I share my thoughts on which nuclear players get hit hardest.

Symptoms of an AI Bubble

Frontier labs have been reporting annualized run-rate revenues of $25 to $30 billion. These murky figures are calculated by multiplying a single strong month by twelve, often at a moment when a product launch like Claude Code can distort numbers well beyond their representative averages. Even accepting those numbers, they sit against enormous training and inference costs, in the case of OpenAI, roughly $600 billion through 2030. All of these compute costs pale in comparison to the supporting infrastructure investments in chips, server racks, datacentres, power and cooling which McKinsey projects to reach 6.7 trillion in capital expenditure by 2030.

Three years after 100 million users were collectively wowed by the Eliza effect of a purposefully designed human-like GPT3 chatbot, subsidized use is still the rule. It is the same playbook Silicon Valley used to underwrite cheap Uber rides and discounted DoorDash deliveries to upend traditional incumbents and establish market share. Sam Altman acknowledged in January 2025 that OpenAI was losing money on every $200/month ChatGPT Pro subscriber, with heavy users consuming an estimated $5,000 in compute: a 25-to-1 subsidy ratio.

The glitch with the LLM subsidy strategy is that the per-user cost of frontier AI is an eye wateringly expensive data centre, not a low wage driver’s time and unlike taxi rides and takeout food no one has yet demonstrated that the users who showed up for the subsidized LLM product will follow it to true market price.

In the absence of organic revenue what is propping up these investments is in part what technology analyst Azeem Azhar has called a “financial ouroboros:” the same capital cycling through chipmakers, cloud providers, and frontier labs in a closed loop. A single point of demand failure could now propagate through every balance sheet in the chain simultaneously.

The agentic era was supposed to be where the economics finally worked. However, early deployments are raising questions. Uber deployed Claude Code to roughly 5,000 engineers in December 2025 and exhausted its entire 2026 AI token budget by April, with heavy users burning $500 to $2,000 per month. The COO subsequently admitted the company could not draw a clear line between that spending and improvements in consumer products. Global research and advisory firm Gartner projects that enterprise AI bills will rise through 2030 even as per-token costs fall, because agentic workflows consume between 5 and 30 times more tokens per task than standard chatbot interactions. While still early in the technology’s deployment, a July 2025 MIT study of 300 enterprise AI deployments found that 95% delivered no measurable P&L impact.

Willingness to pay significantly more depends on products which are truly economically transformative. The obvious prize is white-collar labour displacement, a 5 to 10 trillion dollar opportunity. However, the predictions of Microsofts AI CEO, Mustafa Suleyman, in February of this year that most tasks that involve “sitting down at a computer” will be fully automated by AI within the next year or 18 months is looking increasingly outlandish. While there has been a junior white-collar slowdown brewing over the last few years it is largely explained by the extraordinary numbers of people hired by tech companies between 2020 and 2022 on the assumption that pandemic-era growth was permanent. ‘We are restructuring toward an AI-first workforce’ lands better with investors than ‘we hired 40,000 people we did not need during a zero interest rate frenzy,’ and CFOs have not been shy about exploiting the ambiguity.

LLM architecture a dead end?

Despite our hardwired tendency to anthropomorphize coherent language production stretching all the way back to Joseph Weizenbaum’s primitive ELIZA chatbot, Large language models are at their core next-token predictors: highly sophisticated auto-complete. They generate confident outputs that sound correct by pattern-matching against a vast training corpus but outside of domains like coding and mathematics, where outputs can be tested against an external ground truth, the model has no equivalent check on whether a fluent sentence happens to be a true one.

The resulting hallucinations and reliability problems appear not to be a temporary deficiency that more training data or compute will resolve. Initially scaling impressed researchers with emergent properties like multi-step reasoning and in-context learning such that it was informally codified into a scaling law. However the phenomenon seems to have plateaued with focus shifting to post training optimization. OpenAI cofounder Ilya Sutstever has stated that the age of scaling is ending and we’re back in the age of research. Yann LeCun, who spent 12 years as Meta’s chief AI scientist, departed in late 2025 to found AMI Labs on the explicit thesis that hallucinations are intrinsic to the LLM architecture rather than fixable with more data. “LLMs are too limiting,” LeCun stated. “Scaling them up will not allow us to reach AGI.”

These comments should raise skepticism about all in bets on a massive infrastructure and compute spend to support a technology that may not be able to meet expectations sufficiently to earn the revenues necessary to pay back the extraordinary investments.

The Fallout of an AI meltdown

If the AI bubble deflates, the nuclear sector absorbs collateral damage in proportion to how closely a given project was tethered to that capital.

Uranium fundamentals remain intact regardless of what happens to Nvidia’s valuation. Commercial inventories are thin. The secondary supply sources that suppressed prices for 30 years are gone. A projected gap of more than 1 billion pounds between utility requirements and currently visible supply over the next 20 years looms over slow but steady demand growth from China, India, and Russia’s export program.

A speculative correction will likely push uranium spot prices down, as financial buyers exit and risk-off sentiment dominates. This however could worsen the structural deficit heading into the 2030s as lower prices could delay investment in new mines that must come online quickly to narrow the supply gap.

The advanced reactor companies that tech giants chose as partners occupy a more precarious position. The logic behind those choices reflected what might charitably be called disruption bias: a narrative driven tendency to reach for what are being portrayed as “advanced” reactors held back by overzealous regulation while an inferior conventional reactor fleet failed to make good on the promise of electricity too cheap to meter.

Each tech giant has chosen its own distinct advanced reactor darling, a pot pouri of technologies that have frustrated previous efforts to commercialize them. Meta chose the sodium fast reactor vendor Oklo, Google chose a TRISO fueled molten salt cooled reactor concept being developed by Kairos and Amazon chose the recently IPO’d pebble bed high temperature gas reactor start-up X-Energy.

All of these companies are in the protoype stages. None have a licensed, operating commercial reactor design. Each one is asking its “Magnificent Seven” patron to co-finance the creation of industrial supply chains, regulatory precedents, and fuel fabrication infrastructure that does not yet exist at commercial scale.

Uprates are the most defensible position in a correction scenario. The construction risk is lower, the grid interconnection is already built, and the incremental capacity comes online on a timeline measured in years rather than decades. See my interview below with James Krellenstein and Robb Steward of Alva Energy Inc. for the bulls case for this kind of “boring nuclear” additions.

If demand projections are revised significantly downward as the AI data centre buildout slows, stranded baseload generation becomes a genuine utility-sector fear and the business case for large new builds weakens materially. Demand certainty is the foundation beneath every capital-intensive, long-duration, high-risk energy project, and that AI driven demand certainty looked considerably more solid in 2023-2025 than it does today.

The projects most likely to survive a bursting AI bubble are the ones that were least dependent on AI demand projections or venture-capital patience: refurbishments, uprates, and the least technologically speculative new build reactors, those that already have a licensed design, an operational history and a utility customer with growing demand not solely dependent on AI datacentres.

Without a genuinely transformative AI product that can command prices reflecting actual compute costs, the capital cycling through chipmakers, hyperscalers, and frontier labs has no exit other than a correction. Nuclear investors who have anchored their business cases to data centre demand projections made at the peak of that cycle should be doing the math on what their projects look like if those projections are halved.

Like this post. Tokens of appreciation are free and help us to bring you more great content.

The Book of Mirdad by Mikhail Naimy

I have another life. Outside of energy journalism and advocacy I am an ER doctor. The other night, I had an interaction with an elderly patient that unexpectedly brought my two lives together. He was in his late 70s. He’d lost his house, was living out of his car and was drinking more than he should. He was in rough shape.

I asked him what he used to do for a living.

“Draftsman at Ontario Hydro.”

I probed further. Turns out his career included drawing up blueprints for every one of Ontario’s nuclear power plants. I told him I was interested in nuclear power; he understandably but rather depressingly interpreted “interested” as “anti”.

“Yeah, a lot of people don’t like it,” he said.

I clarified my position, thanking him for being part of the great generation who’d helped to build the infrastructure that underpins our health and prosperity to this day.

He was taken aback. It took a few seconds but his eyes lit up with pride and a sense of rediscovered dignity that shone through his matted hair and unwashed clothes as he lay vulnerable on that emergency room stretcher.

Subscribe now

Before Computer Aided Design (CAD) arrived in the 1980s, draftsmen stood at drawing boards and used pencils, compasses, rulers, and protractors to translate engineering intent into buildable reality. They specified not just what each system needed to do but exactly where every pipe run, cable tray and penetration would sit in three-dimensional space, how it would connect to everything around it, what clearances it needed and how a tradesman standing in front of it with tools in hand would actually build it.

The drawings conducted a symphony of mechanical, electrical, and civil disciplines. If two systems conflicted, someone who understood both the design intent and the physical reality had to catch it at the drawing board before it became an expensive problem in the field.

The 16 reactors that now supply roughly half of Ontario’s electricity were built from hundreds of thousands of such drawings.

My generation struggles to build even modest infrastructure projects: The Toronto Eglinton Crosstown Light Rail Project began in 2011 and only went into service in February of this year. A series of technical problems and associated cost overruns meant that a 19-kilometre light rail line took 15 years to open. We seem to have forgotten how to build the infrastructure we need to renew and grow our cities.

In contrast our parents’ and grandparents’ generations performed miracles. Between 1966 and 1993 Ontario’s power demand nearly tripled, from 51 TWh to 135 TWh. Engineers and tradespeople managed to keep up by constructing 20 large CANDU reactors between Pickering, Darlington and Bruce nuclear generating stations, all in just 27 years. In the current moment we look to China with a mixture of awe and unease as they build highways, bridges and power plants at what feels like lightning speed. We once moved at that pace.

To visit the nuclear stations drafted by my patient, as I have done over the last 8 years, is to be struck both by their monumental size and the sheer volume of electricity they generate. Today these reactors produce power sufficient for 10 million households, all from four buildings the size of modest shopping malls. This energy base has underpinned our productive economy, enabling booming auto and steel sectors and the manufacturing base of Canada’s most industrialized province.

In that ER hallway, speaking about more than just his worn down body, I saw reflected in this man’s aging eyes the harsh hospital lights powered mostly by nuclear energy: literally, the fruits of his labour.

Our society seems finally to have reached a political consensus that we should not be tearing down what my friend and energy historian Emmet Penney refers to as our industrial cathedrals built for us by this great generation: edifices like our nuclear power plants at Pickering, Darlington and Bruce.

Share

The champions of that kind of vandalism have been decisively defeated. Just this week Avi Lewis, the newly crowned Leader of Canada’s left wing New Democratic Party, once the historic party of labour representing the interests of workers like my patient, now the party of urban elites, renounced his previously firmly held position favouring nuclear phaseout.

I’m glad to have played my role in these battles as co-founder and President of Canadians for Nuclear Energy. From 2020-2024 against what we were told were impossible odds, we led the successful campaign to save and refurbish the 2000 MW Pickering B nuclear station, which was drafted by my patient and had been slated for decommissioning in 2025. The result? The preservation of 7200 direct and indirect jobs and 15% of Ontario’s electricity production for another 30 years. The power plant will last at least 80 years and outlive the now grey haired environmentalists who made it their live’s work to tear it down.

It was an honour to provide clinical care for this man and satisfying that in my strange moonlighting career as a nuclear energy advocate I had also cared for and preserved his legacy.

As I grow into my mid forties and spend the best parts of my days trying to instill the values and lessons I’ve gathered into my 7 year-old son, I have become more sentimental when thinking of my elders and their contributions. Beyond displays of gratitude we owe them something deeper.

It starts with ensuring that our children and grandchildren understand and steward their inheritance, never take it for granted and draw inspiration to pick up their tools to leave their mark, continuing our never ending struggle against entropy and decay, to build healthy, wealthy societies worthy of our ancestors’ toil.

Like this post and share it, consider it a blueprint for intergenerational gratitude.

Share

The gas turbine emerged only in the 1930s. It is the most recent and the last prime mover. It has subsequently become the undisputed champion of global aviation and increasingly a dominant source of dispatchable electricity on grids around the world. Demand for gas turbines is now surging as AI data centres require rapid additions of baseload power.

John Barber patented a recognizable gas turbine cycle in 1791, describing a compressor, a combustion chamber, and a turbine wheel in sequence. John Brayton worked out the thermodynamic cycle that still bears his name in 1872. By the time Frank Whittle ran his first jet engine prototype in 1937, engineers had been sketching versions of the same concept for 65 years.

A gas turbine is, at its core, a device that compresses air, adds fuel and combusts it to produce extremely hot gas to spin a turbine. As engineers push the thermodynamic limits through hotter temperatures the engineering reality is that the hot gasses leaving a modern combustor are well beyond the melting point of the alloys from which the turbine blades are made.

The compressor in a large commercial engine reaches pressure ratios above 40:1, requiring blade geometries machined to tolerances measured in microns across components spinning at thousands of revolutions per minute. In aviation applications the containment structure must be light enough to fly and strong enough to survive a blade release without sending a fragment through a fuselage.

The extreme safety and reliability requirements in aviation and power generation impose a slower pace in innovation cycles reminiscent of the nuclear industry. In the end the gas turbine’s limits are defined by the cumulative materials and manufacturing knowledge required to operate at the edge of its thermodynamics. These factors determine who can build it, and why it is one of the few technologies that the West still decisively dominates, for now.

The following essay is based on a conversation on the Decouple Podcast with Dr. David Helmer, a veteran of GE Global Research and the American Society of Mechanical Engineers (ASME) K-14 gas turbine heat transfer committee.

The impetus for the development of the gas turbine

What finally shifted the gas turbine from blueprints to a functioning engine was the pressures of a World War in which aviation had become increasingly central. Through the 1930s, piston engines remained the only practical option for powered flight, and their limitations were becoming a military liability.

Piston engines struggle at higher altitudes, where thin air starves combustion and power drops sharply. They are mechanically complex, with hundreds of reciprocating and rotating parts subject to vibration, wear, and failure. Their power-to-weight ratio, while adequate for the propeller aircraft of the era, left little room for the performance gains that military planners were demanding: greater speed, higher altitude, longer range. Frank Whittle in Britain and Hans von Ohain both recognized that the piston engine had reached a ceiling that the Brayton cycle could break through.

The early engines were extraordinary in concept and crude in execution. Whittle’s W.1, which powered the Gloster E.28/39 on its first flight in May 1941, produced 860 pounds of thrust and had a service life measured in dozens of hours before its primitive nickel alloy components failed. The materials problem, which would come to define the entire subsequent history of the technology, had to wait.

What the war years established, beyond the basic feasibility of jet propulsion, was the industrial and institutional foundation that would separate the countries that could build serious gas turbines from those that could not. Britain transferred its early jet technology to the United States during the war which seeded GE and Pratt & Whitney’s programs and gave American manufacturers a running start they have still not yet relinquished. Germany’s early expertise withered due to the destruction of industrial infrastructure, postwar restrictions on German military aviation and the migration of engineers into allied programs after the war. They only recovered in 1960-70’s but this time in the heavy frame gas turbine power generation market.

The know-how embedded in these technology transfers was only partly captured in drawings and specifications. Much of it lived in the hands and judgment of engineers and machinists as tacit knowledge.

The material science miracles of the modern gas turbine

The central engineering challenge of the gas turbine is that it operates in conditions that should destroy it. Hot gas leaving a modern combustor reaches around 1,500 degrees Celsius. The nickel superalloys used in turbine blades begin to lose structural integrity at temperatures several hundred degrees below that. Engineers have spent decades developing systems to keep them alive: internal cooling channels machined into the single crystal turbine blade itself, thermal barrier coatings and the superalloys doped with rhenium that extend the temperature range the metal can tolerate before creep and fatigue set in.

The design philosophy imposed by the extreme conditions within the engine has no real parallel in other high-stakes engineering domains. A turbine blade in service is always degrading or in the parlance of the industry “experiencing distress.” It cannot however fail catastrophically. The first turbine stage can be “mostly missing,” burned back by inadequate cooling or a material defect, and the engine will often still fly for many more flight cycles until measured degradation crosses a defined tipping point. The system is designed to tolerate continuous partial failure of its most stressed components because pushing to the edge of material limits is the only way to achieve the power density, fuel efficiency and longevity that modern aviation, and power generation demand.

This approach, coupled with the incredibly high safety stakes involved in commercial aviation, requires extensive evaluation. That risk management focus imposes a conservative innovation cycle well familiar to those knowledgeable about nuclear power. A modification to a commercial jet engine requires bench testing of the component in isolation, then stand testing under representative operating conditions which sometimes includes the rigors of the famous chicken gun and blade out scenario. Next is flight and endurance testing for hours accumulation, and finally aircraft recertification if the change affects airframe performance.

The gap between the GE90 and its successor the GE9X is roughly 25 years, a near irreducible time required to validate changes in a system where GE applies 6 sigma quality standards across its engine programs or fewer than 3.4 defects per million.

For reference SpaceX which has moved at lightning speeds in its rocket development is able to accept an engine failure rate of roughly 1% across its launch program.

Gas Turbines go Terrestrial

When a gas turbine leaves the aircraft and gets bolted to the ground, some of the engineering constraints relax. Weight stops mattering and bird ingestion becomes a smaller problem but precision engineering around performance and reliability still matters. Unlike a jet with relatively short duty cycles, a utility turbine may be asked to run continuously for weeks, while every small loss in efficiency compounds into fuel cost and every excess maintenance outage becomes lost revenue.

The most interesting design option that became available to terrestrial gas turbines, however, was the potential to use their extremely high quality waste heat which averages 550 degrees Celsius. A gas turbine in simple cycle operation, running alone without any additional heat recovery, converts fuel to electricity at thermal efficiency in the low to mid 40s percent. Unlike an aircraft, where that heat is simply dumped into the ambient air, on the ground it can be routed into a heat recovery steam generator, which uses it to boil water and drive a steam turbine in a second, separate power generation cycle.

Early combined cycle plants existed decades before, but the concept only became commercially transformative in the 1980s and 1990s as improved materials, higher turbine inlet temperatures, blade cooling, digital controls, and increasingly sophisticated combustion management pushed gas turbine performance high enough to make the economics compelling. Combined cycle plants could now exceed 50 percent thermal efficiency and eventually move past 60 percent, allowing natural gas to compete directly with coal for sustained baseload electricity generation rather than merely serving as an expensive peaking fuel.

At the same time, deregulated electricity markets were expanding across North America and Europe, rewarding technologies with lower upfront capital costs, shorter construction schedules, and operational flexibility. Combined cycle gas turbines fit that environment unusually well.

They were efficient enough to run continuously, could ramp up and shut down far more easily than coal or nuclear, and avoided fuel costs entirely when offline, an important advantage in volatile merchant power markets. Environmental pressure on coal was also rising due to concerns over sulfur dioxide, particulates, mercury, and eventually CO2 emissions, areas in which natural gas is a demonstrably cleaner alternative to coal.

GE, Siemens Energy, and Mitsubishi Heavy Industries spent the late 1990s and early 2000s expanding manufacturing capacity aggressively around the assumption that the world was moving toward a gas dominated electricity system built on combined cycle gas turbines. Annual heavy duty gas turbine orders reportedly reached 120 to 140 gigawatts globally in the early 2000s.

Then a perfect storm hit. The 2008 financial crisis crushed electricity demand growth across much of the OECD, wind and solar rapidly absorbed incremental generation growth and the merchant power boom had already saturated the American market with relatively new gas plants. Global heavy-duty turbine orders fell to roughly 30 to 40 gigawatts annually by the mid-2010s, a collapse of more than two-thirds from peak.

The consequences were severe: massive layoffs, factory underutilization, write-downs, collapsing margins, and an investor confidence collapse that fed the broader GE crisis of the late 2010s. Under Larry Culp, GE eventually broke itself into 3 separate companies: GE HealthCare, GE Vernova, and GE Aerospace.

Today GE Vernova, Siemens Energy, and Mitsubishi face another major positive demand shock driven by AI data centers and electrification, with delivery queues stretching toward the early 2030s. This time none of them are expanding recklessly recognizing the possibility of an AI bubble, a strategic stance somewhat akin to the newfound capital discipline in the U.S. shale patch that evolved after its most recent price collapse.

The gap is being filled by companies like FTAI power which is currently taking CFM56 engines off retired commercial aircraft, converting them for ground-based power generation, and selling that capacity to data centers that cannot wait for purpose-built equipment.

Why the West is Still Dominant in Gas Turbines

Commercial jet engine manufacturing is one of the most concentrated industries on earth. GE Aerospace, Rolls-Royce, Pratt & Whitney and France’s Safran produce essentially all of the turbofan engines powering large commercial aircraft.

Fighter jet engines operate under a different set of constraints: shorter service lives, no six sigma safety requirement, and performance prioritized over durability. Both Russia and China have developed credible military engines on that basis, but the jump from a capable fighter engine to a certified commercial turbofan is a different engineering problem entirely.

So why has that dominance proven so durable despite decades of well-resourced attempts to break it? The answer has several mutually reinforcing layers.

First is tacit knowledge; the accumulated manufacturing process understanding, inspection infrastructure, and decades of operating data required to demonstrate 6 sigma reliability simply cannot be extracted from a blueprint or a reverse-engineered component.

Secondly, certified batches of the doped superalloys required for hot section components are sourced from a small number of qualified suppliers whose relationships with the leading manufacturers took decades to build.

Lastly, the precision machine tools required to produce single crystal turbine blades to micron tolerances are themselves a specialized product made by only a handful of manufacturers, with German toolmakers historically prominent among them.

Behind it all sits a century of accumulated operating hours and failure data that the leading manufacturers have built into their design and certification processes. The gap China, for example, faces is having to reconstruct every one of those layers simultaneously.

The terrestrial gas turbines market is similarly concentrated in the Western bloc. Siemens Energy and Mitsubishi Power compete effectively in the heavy frame market alongside GE. The materials demands are less extreme and the certification requirements less stringent but neither Russia nor China has yet produced a competitive large combined cycle turbine without technology transfer or joint venture arrangements with one of the established Western or Japanese leaders.

Russias Gas Turbine woes

The Russian case illuminates what happens when a high-performance industrial capability like gas turbines is simultaneously hit from multiple directions. The deterioration predates the 2022 invasion of Ukraine by decades, beginning with the Soviet collapse and the economic chaos that followed. The engineering talent base that had sustained Soviet aviation and power generation programs dispersed, with experienced engineers leaving for better-paid work abroad or moving into unrelated sectors as defense and industrial budgets collapsed.

The materials problem compounded the talent problem. Russia’s ability to maintain supply relationships for specialized parts, never strong by Western standards, eroded through the 1990s and 2000s as the industrial base contracted.

Then came the precision machining problem. Germany was Russia’s primary supplier of the high-precision machine tools required to produce turbine components to the tolerances modern engines demand. That relationship was embedded in a broader economic interdependence: Russian gas flowing west, German industrial equipment flowing east. Post-2022 sanctions put an end to that special Ost-Politik relationship.

GE’s withdrawal from Russian power generation maintenance under threat of sanctions compounded the crisis at the operational level. GE had engineers on the ground maintaining the most capable turbines running on the Russian grid, many of them originally supplied through the Siemens joint venture arrangements. When those engineers left, the knowledge and tooling they brought with them also disappeared.

The Russian grid is now running aging equipment with a degraded maintenance capability, shortening component life further and creating a compounding reliability problem. Recent adaptations by Russia to overcome sanctions pressure include turning to Iran which has a small but self sufficient gas turbine industry.

The West still dominates. When will the rest catch up?

The gas turbine waited nearly 150 years from theory to practice to join the pantheon of prime movers because the materials and manufacturing precision it required didn’t exist. Its arrival reorganized the world: global aviation, just-in-time supply chains built on air freight, and the most efficient heat engine ever built for generating electricity.

That the Western bloc and Japan still decisively control this technology after nearly a century is the outcome of compounding advantages in materials science, precision manufacturing, tacit institutional knowledge, and certified failure data accumulated across millions of operating hours.

China’s “Two Engines” programme, launched in 2013 and backed by sustained state investment, has not yet closed that gap in either commercial aviation or heavy-frame combined cycle turbines. Whether it eventually does will matter as much to the future balance of industrial power as it does for the semiconductor or battery supply chain contest currently dominating policy attention.

The AI demand surge has arrived into this constrained supply environment with considerable force. CCGT plant costs have roughly tripled over the last 3 years, order books stretch toward the early 2030s, and the gap is being filled in part with jet engines from retired aircraft converted for ground-based power generation. Whether the demand wave will persist long enough to justify genuine capacity expansion, or correct as the 1990s boom did, is the question that manufacturers must judge.

Gas turbines represent an accumulated industrial inheritance which took nearly a century to come within small percentages of theoretical maximum efficiencies. The queue is long in part because the capabilities of those who have mastered the last prime mover are exceptionally rare.

For more on David Helmer visit his website https://www.nttwconsulting.com/

Like this post to keep us pushing the edge of our material limits.

The first part of this analysis published last week established why uranium behaves unlike any other commodity: no real-time price discovery, demand that arrives in waves separated by years of buyer silence, and a cyclical pattern sustained across 6 decades by a secondary supply wedge that has now been exhausted.

Today we examine both sides of the ledger simultaneously, because the supply story and the demand story are not independent.

They are pressing against each other in ways the aggregate global supply-demand charts do not capture. The geopolitical fracturing of the market means the relevant question is no longer whether there is enough uranium in the world. There probably is. The question is whether there is enough in the western supply chain, and that is a considerably more uncomfortable place to look

.

What’s structurally different on the supply side

The absence of secondary supply changes the mathematics of the uranium market in a way that has not yet been fully priced into conventional analysis. When the next demand surge arrives, there is no inventory wedge to release.

The price signal that would previously have been dampened by a flood of government stockpile material or down-blended warheads will now have to travel all the way to primary producers, who will have to respond by sinking capital into new mines that take years to permit, build, and bring to full production. That lag between price signal and supply response is the structural feature that distinguishes this cycle from every predecessor.

Cameco’s position within that structure is deliberate. The company still has 30% of its licensed, permitted production capacity either in care and maintenance or operating below full rate, held in reserve explicitly because sufficient long-term contracting has not yet arrived to justify restart. Isaac was direct about this: there is no build-it-and-they-will-come model in uranium.

Producers commit capital after securing forward demand, not before. Cameco’s Millennium deposit and other Athabasca-adjacent assets remain in the ground as inventory, pounds with a balance sheet value that appreciates as the price environment improves, available for development when the contracting signal justifies it. The discipline is a deliberate strategy of matching production decisions to marketing outcomes rather than running mines at capacity and hoping the market absorbs the output.

Kazatomprom has arrived at the same position by a harder road. Having learned during the post-Fukushima bear that maximizing volume destroys the value of the national asset it is supposed to steward, Kazakhstan’s state producer now runs a contracting-first strategy that mirrors Cameco’s. The production target reductions that initially looked like a sulfuric acid supply problem, output guidance cut 10 to 15% at first and now closer to 25%, are better understood as a capital allocation decision dressed in operational clothing.

The sulfuric acid shortage is real with phosphate fertilizer production competing for available supply, and the constraint genuinely biting into production volumes that Kazatomprom had already sold forward. Building a new sulfuric acid plant takes 18 to 24 months and costs $200 to $300 million. Kazatomprom has done exactly this before and the engineering knowledge exists.

The decision to fund and begin construction came as contracting conditions improved through 2023, not at the moment the shortage first appeared. The plant is now under construction and expected online in 2026. That it was conceived and financed alongside a strengthening contract book, rather than purely in response to an operational emergency, is the tell.

What settles the supply discipline question is Kazatomprom’s 2026 guidance, which has nothing to do with acid. The company announced a further roughly 10% production cut for 2026, stating explicitly that current prices and the existing level of uncovered demand are not sufficient to incentivise a return to full output.

New Athabasca Basin development offers genuine upside against this tightening but carries first-of-a-kind risk that promotional materials for junior developers tend to understate. NextGen Energy’s Rook One project anticipates production of around 30 million pounds per year, which would be transformative for the western supply picture if it materializes on schedule.

But Cigar Lake’s mining method was built on decades of institutional accumulation by a company that had already operated McArthur River and learned the hard way what fractured sandstone and unconformity-hosted ore bodies do when water management fails. The freeze engineering, the jet-boring system, the radiation handling protocols for 15% ore grades: none of this existed as received knowledge before Cameco developed it.

A new operator in a different geological setting, without that institutional base, faces engineering challenges with no established playbook. Isaac described the dynamic precisely: junior uranium projects tend to attract elevated valuations before they have entered what he called the valley of development, the long and capital-intensive period between exploration promise and operational reality.

Cameco’s own reserve and resource pipeline contains around one billion pounds, all adjacent to existing infrastructure and already on the balance sheet. Any external acquisition has to clear that internal hurdle rate. At current junior valuations, very few external projects do, and the moment when they will is the moment those projects have moved through the valley and been repriced accordingly. There is a significant incumbent advantage.

The geopolitical dimension of the supply picture adds a further layer of tightening, one that aggregate global supply-demand charts tend to obscure entirely. Uranium has always been a highly trade-dependent commodity; for most of the industry’s history, around 90% of uranium production came from countries with no civilian nuclear power program, while around 90% of demand came from countries with little or no domestic uranium production. The system worked because no one worried much about where the uranium came from. The unipolar exception to the rule made that comfort possible. Emerging multipolarity is challenging what in retrospect was a naive assumption.

Share

Multipolarity and a fracturing market

The standard supply-demand analysis of the uranium market operates at the global aggregate level: total pounds produced, total pounds consumed, projected deficit or surplus. That framing was adequate for a world in which uranium flowed freely across borders toward whoever offered the best price and utilities had no preference for origin. That world is gone, and the aggregate framing now obscures more than it reveals.

The uranium market is splitting into two increasingly separate procurement spheres. On one side sit US, European, Japanese, and South Korean utilities buying uranium that can pass ESG screens, government security reviews, and shareholder scrutiny.

On the other sit Chinese state enterprises and Rosatom export clients building reactors and fueling them through supply chains largely invisible to western price reporting and largely inaccessible to western buyers. The question that matters for uranium investors is not whether there is enough uranium in the world. There probably is. The question is whether there is enough in the western supply chain, and that is an increasingly different and more uncomfortable question.

China’s nuclear ambitions alone reframe the demand picture in ways the aggregate numbers do not capture. China currently operates around 62 reactors and has roughly 34 more under construction, with stated government targets pointing toward 150 GW of installed capacity by 2035, roughly triple today’s fleet. That uranium demand is being served through a supply architecture running parallel to the western term market: bilateral offtake agreements with Kazakhstan and Uzbekistan, African production assets acquired by Chinese state enterprises, and processing arrangements that often route material through Russian infrastructure. None of this volume shows up in western spot or term prices as demand. It disappears from the pool of available western-origin supply without generating a visible price signal in the markets western investors watch.

The ownership map of global uranium production makes this dynamic more acute than the geography suggests. Kazatomprom is state-owned. Uranium One, which holds significant Kazakh production stakes, is owned by Rosatom. CNNC controls Rössing in Namibia, with output largely committed to Chinese offtake, and Husab, also among Namibia’s larger producers, is similarly Chinese-controlled.

Namibia looks like a western-friendly source until that ownership layer is examined. Niger’s Somair mine was effectively removed from the western supply chain by the 2023 coup. The geographic map of uranium production and the geopolitical map of uranium availability are not the same map, and confusing them produces dangerously optimistic supply assessments.

The spot and term prices that western investors watch are prices in the western procurement sphere. As western-origin supply tightens relative to western demand, those prices will increasingly reflect western-origin scarcity rather than global aggregate adequacy. The fracture is not symmetric, and the side bearing the greater adjustment cost is the side that spent the post-Cold War decades assuming global commodity markets were a permanent and neutral feature of the energy landscape.

What’s structurally different on the demand side

The demand transformation driving this bull has two distinct layers that are worth separating before examining how they interact. The first is the reversal of the post-Fukushima western phaseout. The second is a class of demand that did not exist in any previous uranium cycle. Together they are pressing against a supply base that is simultaneously tightening for the structural reasons described in the previous section.

Russia’s invasion of Ukraine in February 2022 was the demand-side inflection point. Before the invasion, energy security was a minor consideration in power planning; after it, energy security became the dominant concern. Nuclear’s profile changed accordingly. A technology that was sunsetting as a legacy baseload asset in increasingly deregulated western grids was suddenly being re-examined as a mature, dispatchable, carbon-free power source that does not depend on fuel imports from geopolitically unreliable suppliers.

The irony that Russian enrichment services remained deeply embedded in western fuel chains long after the invasion made the point more viscerally than any policy paper could. Belgium reversed its phaseout. Japan began restarting reactors that had sat idle since 2011, with a government target of returning 30 or more units to service. The United States passed legislation extending production tax credits for existing nuclear plants and strengthening support for new construction, while several European governments that had treated nuclear as a managed decline reclassified it as critical infrastructure.

The reactor life extension wave matters enormously for uranium demand because it converts what had been scheduled demand destruction into sustained long-run consumption. A reactor expected to retire in 2030 that instead receives a subsequent license renewal to 2050 adds 20 years of uranium conversion, enrichment, and fabrication demand. Given the relatively small fleet of global reactors, each reactor is individually responsible for roughly 0.2% of global demand. Multiplied across the US fleet, where subsequent license renewals are now the norm rather than the exception, and across Europe and Japan, the aggregate addition is substantial.

Roughly 70 reactors are currently under construction globally, a figure not seen since the 1970s construction wave, and the uranium demand from those reactors is being contracted years before the plants begin operating. The contracting signal is already building in the term market even though the physical consumption is still years away.

Then there is the data center and artificial intelligence demand signal, which did not feature in uranium investment theses as recently as three years ago. Microsoft’s agreement with Constellation Energy to restart Three Mile Island Unit 1, Amazon and Google signing nuclear power purchase agreements, and a broadening field of technology companies explicitly seeking firm 24-hour carbon-free baseload to power compute infrastructure have introduced a class of buyer into the nuclear power market with no historical precedent.

These buyers carry deep balance sheets and a demonstrated willingness to sign long-duration contracts to secure power supply certainty, and their demand is structurally incompatible with intermittent generation: a data center cannot tolerate the output variability of wind and solar without storage infrastructure that does not yet exist at the required scale and cost. Nuclear is the only mature technology that satisfies both the carbon and the reliability constraint simultaneously. The data center signal is still early but it is layering onto an already tightening picture rather than driving it independently.

The procurement sequencing laid out in Section 1 gives a precise read on where western utilities currently sit within all of this. Conversion prices are at historic highs; the western conversion market, dominated by Cameco’s Port Hope facility and Orano’s Metropolis Works, has essentially no spare capacity and is fully contracted well into the next decade.

Enrichment is similarly tight: after the effective sanctioning of TENEX, Russia’s enrichment export arm, western utilities were left with Urenco and Orano as their primary options, both running at high utilization, and new enrichment capacity takes 10 or more years to permit, finance, and construct. Fabrication contracted, enrichment contracted, conversion contracted. The uranium contracting wave has not yet fully arrived. When it does, it will hit a primary supply base deliberately held in partial restraint, with no secondary supply available to buffer the price signal.

What Cameco’s Vertical Integration Signals

Cameco’s growing vertical integration spans the full western nuclear fuel cycle by design. The 1988 merger of two Canadian Crown corporations created a company that ran from Athabasca Basin mining through milling, refining, and CANDU fuel fabrication from the outset. What has changed over the subsequent three decades is the deliberate extension of that integration into the light water reactor fuel cycle, which accounts for roughly 92% of the world’s operating reactors, culminating in the Westinghouse acquisition completed with Brookfield Asset Management in 2023.

That acquisition is best understood as demand creation. Between what Cameco handles for pressurized heavy water reactors and what Westinghouse handles for light water reactors, around 65% of western world fuel bundles pass through a facility owned by one of the two companies. Every AP1000 that Westinghouse sells and builds is 80 to 100 years of upstream demand for uranium, conversion, enrichment, and fabrication. The logic is circular in the most productive sense: building reactors creates the long-run demand that justifies investing in the upstream supply chain to serve them.

In the fracturing market described in the previous section, Cameco’s ownership structure is itself a competitive asset in a way it was not a decade ago. Western utilities under government pressure to demonstrate supply chain security are not indifferent to the fact that a long-term contract with Cameco is a contract with a Canadian-domiciled, privately owned company whose assets sit in Saskatchewan and Ontario, jurisdictions with stable rule of law and no plausible scenario in which production gets diverted to a state strategic priority. That provenance premium is real and growing as geopolitical sorting accelerates.

Cameco’s Athabasca Basin operations rest on institutional knowledge that cannot be acquired quickly regardless of capital availability: freeze engineering, jet-boring protocols, radiation management at ore grades with no parallel elsewhere in the western world. That knowledge lives in a small workforce operating in one of the most remote industrial environments on the planet, and it is not easily replicated. NexGen and Denison are competing for the same scarce pool of experienced operators and engineers that Cameco has spent decades cultivating.

The labour market in northern Saskatchewan is not elastic. You cannot import this expertise from a copper mine or a conventional uranium operation in Kazakhstan or Namibia. The tacit knowledge embedded in Athabasca Basin operations is specific to that geology, that ore grade, and those extraction methods, and it accumulates slowly through direct experience rather than capital deployment.

Each new deposit presents its own first-of-a-kind engineering problem: McArthur River and Cigar Lake sit 55 kilometres apart and require fundamentally different extraction strategies, and there is no reason to expect the next generation of deposits present less novel mining engineering challenges.

The physical moat and the human capital moat reinforce each other. These barriers do not appear on a balance sheet, but they are the reason Cameco’s assets command the strategic premium they do, and they are the standard against which junior developer promises should be measured.

The company with the deepest operational knowledge of western uranium supply and the most granular view into western utility contracting is the one holding 30% of its licensed capacity in reserve, waiting for the long-term contracting signal to fully arrive. That patience is the clearest available signal that the uranium bull has not yet reached the point at which the smart money feels compelled to rush.

A more boring bull market, maybe

The long-term uranium contract price has been moving steadily upward for three years without a significant reversal, grinding methodically in one direction rather than spiking on a single supply shock or sovereign buying program the way previous bulls did.

That grinding reflects something no previous uranium bull possessed: two major incumbent producers running explicit supply discipline simultaneously, no secondary supply available to buffer a demand surge, a geopolitical environment pushing western utilities toward a narrower pool of origin-appropriate suppliers, and a demand base being rebuilt on foundations more durable than a single policy announcement.

The structural conditions for a disorderly blow-off top, the pattern that ended every previous uranium bull, are less present in this cycle than in any prior one. Kazatomprom now understands that flooding the market destroys the value of a national asset it cannot replace. Cameco has demonstrated through a decade of curtailments and patient contracting that it will not produce into a weak market.

The secondary supply shock absorber that historically amplified both the upswing and the collapse is gone. And the geopolitical fracturing of the market means a Kazakh production surge is less likely to land in the western spot market than it would have in 2012, because the offtake architecture routing Kazakh material toward Chinese and Russian end users is now more contractually entrenched and more politically durable than it was during the last bear.

The residual tail risks are real. A nuclear accident remains the industry’s defining vulnerability, though it would now be in a world where climate security, energy security, and national security have become the explicit pillars of power sector planning in every major western economy.

Governments that reversed phaseout decisions on energy security grounds are less likely to reverse course again, and the generation of reactors now operating embody safety improvements that make a Fukushima-scale event substantially harder to replicate.

The more immediate uncertainty is timing. The uranium market ultimately depends on utilities bringing forward demand, and that timing is not controllable by producers. Conversion and enrichment contracting already reflect the tightened supply picture.

When uranium contracting fully catches up, the price signal will reach producers holding capacity in reserve for exactly that moment. Whether that resolves in an orderly repricing or a violent catch-up move depends on how quickly the last domino falls and how much new primary supply can be brought online in response.

What this cycle has that its predecessors lacked is a structural foundation underneath that uncertainty and that is what makes it worth understanding on its own terms.

Like this post to help balance supply with surging demand for Decouple Media content.

The standard explanation is that nuclear power is back. After a decade of retreating political support following Fukushima, governments across the western world have rediscovered that a 24-hour, carbon-free, energy-dense power source is not something you can replicate with wind turbines and battery storage. Reactor life extensions, phaseout reversals, a Chinese nuclear construction boom, and corporate power purchase agreements from data centre operators hungry for firm baseload have all landed within a compressed window. The demand story is real.

But demand alone does not explain why this bull has behaved so differently from its predecessors. Previous uranium cycles also had credible demand stories. The 2005-2011 bull had a nuclear renaissance narrative, a supply shock from the Cigar Lake mine flooding and China’s sovereign contracting program driving long-term prices. It still collapsed.

Understanding why this cycle may not follow the same script requires understanding something more fundamental: uranium is a commodity unlike any other, governed by market mechanics that have no real parallel in oil, gas, or industrial metals. The history of uranium investing is largely a history of analysts importing mental models from those markets and being destroyed by them.

Grant Isaac, President and COO of Cameco Corp and a former dean of the Edwards School of Business at the University of Saskatchewan, laid out that case in a recent conversation on Decouple. What follows draws on that interview as its primary analytical frame.

Why Uranium is an Unusual Commodity

The single most counterintuitive fact about the uranium market is that it has, in Grant Isaac’s framing, “basically zero fundamental in-year demand.” Every reactor currently operating will load its next fuel bundle from uranium that has already been procured. No utility anywhere on the planet is going into the market today to buy uranium it needs in the next 12 to 18 months. That material is already sitting in a warehouse, in a conversion facility, or moving through the enrichment and fabrication queue.

This inverts every intuition imported from oil markets. When oil demand rises, refiners buy more crude. When natural gas gets tight, utilities burn through spot supply and prices spike within weeks. The feedback loop between physical demand and price is immediate and visible. In uranium, that feedback loop is broken by design.

Utilities treat nuclear fuel as a long-lead procurement item. It is closer in spirit to steam generators or a reactor pressure vessel, ordered years in advance through carefully negotiated bilateral contracts, than buying natural gas. Global uranium demand currently runs around 175 million pounds per year. What passes for a spot market amounts to roughly 50 million pounds traded per year, of which utilities account for perhaps 10 million pounds of discretionary buying: outage adjustments, inventory top-ups, opportunistic purchases when price dips.

The real market is the bilateral term contract market, negotiated out of public view, and the price it produces can only be seen in the rear-view mirror rather than a windshield. It takes a week of negotiations to surface the spot price and a month of transaction volume to establish the term price. There is no real price discovery: no Bloomberg refresh, no live tick, no forward curve with the liquidity of an oil futures market. The uranium price is always, to some degree, a historical artifact.

This structure produces a distinctive and punishing cyclical pattern. Utilities do not bring forward demand to the market on a predictable annual schedule. They can go well below replacement contracting for years, drawing down existing inventory and long-term contract coverage, before returning en masse. Historically when they return, they all tend to arrive simultaneously, and the resulting demand surge hits a supply base that has had no signal to expand.

The uranium cycle runs 10 to 12 years on average, against something like potash’s 18 months, because the inventory buffer is deep and the procurement horizon long. A reactor running today has enough fuel in the core for 18 months and likely has one or two strategic reloads on site. European utilities have historically held roughly two years of forward inventory; Asian utilities typically hold three years or more.

A nuclear plant could theoretically run through a complete fuel cycle, reload from strategic inventory, and run another full cycle without a single uranium delivery arriving at the gate. No other commodity consumer operates with that kind of structural insulation, and it is precisely what enables buyer strikes lasting years and makes uranium demand, when it finally returns, arrive all at once.

Utilities do not contract for uranium in isolation either. They work backward through the fuel cycle: securing fabrication capacity first, then enrichment, then conversion, then uranium. You need to know where the uranium is going before you contract for it. This sequencing means uranium contracting always lags the broader nuclear procurement cycle. When conversion prices are at historic highs and enrichment contracting is strong, as both are right now, uranium demand is still building toward its own contracting surge. The uranium market is, in a structural sense, the last domino.

Share

A Primer on Uranium Production

Not all uranium is created equal, and the method by which it comes out of the ground shapes much of its cost structure, its flexibility as a supply source, and its geopolitical character. There are three primary production methods worth understanding before examining the current supply picture.

Open-pit mining is the simplest and historically the most throttleable: costs load heavily into variable rather than fixed, making it easier to idle when prices fall. Most of Africa’s uranium output has come from open-pit operations. Underground hard-rock mining carries enormous fixed infrastructure costs, which is precisely what makes Canada’s Athabasca Basin’s extraordinary ore grades necessary to justify the investment.

In situ leaching, or ISL, is what the late Bob Steane, Cameco’s longtime chief operating officer, called “plumbing, not mining.” No shaft is sunk, no ground is excavated. Injection wells push a lixiviant, a leaching solution, typically sulfuric acid or caustic soda through a bounded underground sandstone formation. The solution dissolves uranium from the host rock, gets pumped back to surface, and the uranium is precipitated out chemically. Capital requirements are a fraction of a conventional mine, the operation scales by drilling more wells and laying more pipe, and because you are working with permeable sandstone geology rather than hard basement rock, ISL can be economical at ore grades that would send a conventional miner packing.

Kazakhstan’s geology is almost perfectly suited to it: enormous bounded sandstone formations at manageable depths, with none of the freeze challenges or water management complexity that define Athabasca operations, and at a scale that makes every other ISL jurisdiction look like a pilot program. A large US ISR project produces around 200,000 pounds per year against global annual demand of roughly 175 million pounds. Scale is the entire economics of mining, and Kazakhstan has the denominator the US does not.

Then there is the Athabasca Basin. Picture it as a stone bowl: competent granite basement rock forming the bowl itself, fractured sandstone filling it. That sandstone holds enormous volumes of water that communicate across the entire formation through the fractures. The fracturing is also what concentrated the uranium in the first place: Over millions of years, oxidizing groundwater leached uranium from the overlying sandstone and carried it downward in solution. At the unconformity, where those oxidizing waters met reducing conditions in the basement rock, the uranium precipitated chemically and accumulated in extraordinary concentrations. However, the same geology that created the orebody makes extracting it so punishing.

McArthur River sits where a geological anomaly allows most mining to take place in competent basement rock adjacent to the ore. Cameco drives freeze curtains outward from the granite into the surrounding sandstone, creating barriers against water inflow, then uses blasthole stoping with remote-operated scoop trams handling the highest-grade material to manage radiation exposure.

Cigar Lake, just 55 kilometres away, presents a completely different problem: the ore sits at the top of a clay funnel at the unconformity and the geology makes working from above impossible. All operations are conducted below the ore body. Cameco bulk-freezes the entire ore zone from the surface, then bores upward into the frozen material using high-pressure water jets, collecting the resulting uranium slurry through steel pipes feeding into processing circuits below. In both cases without the freeze, the operation does not exist.

The ore grades are what make this extraordinarily complex mining engineering worthwhile. Cigar Lake averages around 15% uranium by mass, having peaked at 18% early in its mine life. McArthur River runs 7 to 8% now. The global average uranium ore grade is under 0.1%. Individual ore passes at Cigar Lake have been recorded at 70% uranium. That grade advantage is the economic justification for every dollar spent on frozen ground, remote operations, and decades of institutional learning about water management in fractured sandstone.

The concentration of production that results is striking. Ghawar, the largest conventional oilfield ever found, supplies around 6% of global oil. Cigar Lake alone supplies 14% of global uranium. McArthur River at full licensed capacity adds another 10 to 13%. Two mines in the same basin in northern Saskatchewan, both accessible only by winter road, account for roughly a quarter of all primary uranium production on earth.

That concentration also explains why the IAEA’s Red Book figure showing Australia with the world’s largest uranium resources is less reassuring than it appears: a resource is uranium believed present based on geological evidence, while a reserve is what remains after asking whether the ore can be extracted at today’s prices and sold at a margin. When that model is applied, the global supply picture narrows considerably. Large numbers in the Red Book are not the same as commercially-available pounds.

A history of uranium cycles

The uranium market has a pattern, and it has repeated with enough consistency that understanding it is the prerequisite for evaluating any claim that this time is different.

Every bull market in uranium history has ended the same way: either a demand destruction shock or a supply flood killed it, and in both cases the ensuing bear market was then sustained and deepened by a vast overhang of secondary supply that took decades to work through. It is the fact that the overhang is now gone that is the structural break. But to appreciate why it matters, you have to see how it functioned across 6 decades of cycles.

The originating myth is Eisenhower’s Atoms for Peace speech and Lewis Strauss’ claim of nuclear power “too cheap to meter.” This seeded an expectation not of hundreds of reactors but thousands.

On the strength of that expectation, uranium was explored for, discovered, and produced at scale well before the demand existed to absorb it. Around 650 reactors eventually came into operation globally, not the 2,000 or more that had been projected, and the surplus from that miscalculation would shadow the uranium market for the next 6 decades as “secondary supply” (defined more broadly as material from stockpiles, military programs, and reprocessing) competed with freshly mined uranium.

Three Mile Island in 1979 and Chernobyl in 1986 compounded a demand collapse that had already begun: reactor orders had peaked in 1974 and were in freefall with more than 120 US reactor orders ultimately cancelled. Public opposition hardened across western Europe, and utilities with contracted material they no longer needed sold into a market with nowhere for the pounds to go.

Then the Cold War ended and a new source of secondary supply entered through the megatons-to-megawatts program, which down-blended 500 metric tonnes of Russian highly enriched uranium into low-enriched uranium and fed it into western civilian power markets over 20 years.

The US Department of Energy sold its own UF6 inventory into the market during the same period. Between military stockpile liquidation and the original overproduction overhang, primary uranium mining was competing against its own past for three decades. Primary production has not matched annual uranium demand in any year for the past 30 years. The market ran on inventory.

The 2000s brought the first bull market that looked like it might escape this pattern. Natural gas prices were rising, George W. Bush was at Calvert Cliffs announcing a nuclear renaissance. Generation III+ reactor designs were moving from paper to licensing and even record breaking construction schedules in Japan’s extraordinary ABWR buildout. Uranium moved from around $10 per pound in 2003 toward a $136 per pound peak in 2007, but the spike was driven less by actual demand contracting than by supply shock anxiety.

In 2003, a water inflow at McArthur River reminded the market that mining has risks that liquidating inventory does not. In 2006, the Cigar Lake development project flooded before it had produced a single pound; it was supposed to come online by 2009, and after the flood some in the market began to wonder whether Cigar Lake was simply unmineable. That fear, not a fundamental shift in contracting, drove the price to its 2007 peak.

Then in the summer of 2010, something demand-driven happened. China arrived in the term market and contracted 152 million pounds of uranium with Cameco, Kazatomprom, and Orano: nearly a year’s worth of total global demand, with deliveries running from 2015 into the early 2020s. Traditional western utilities, watching the scale of sovereign Chinese buying, suddenly recalculated their own forward exposure and tried to go through the contracting door simultaneously. The price was heading toward another high when, in March 2011, a tsunami hit the Japanese coastline and disabled the Fukushima Daiichi plant.

Japan shut down all 50 of its operating reactors, removing roughly 10% of global uranium demand almost overnight. Japan had already contracted years ahead for uranium it now did not need, and that material began finding its way back into the spot market.

Kazatomprom, operating at low marginal cost through ISL and carrying the production volumes it had built up during the bull, kept producing regardless. With no contracting discipline and no floor under which it would reduce output, Kazakhstan jammed surplus pounds through the spot market continuously.

The price collapsed from around $80 per pound to a trough of $17.75, the level at which Cameco finally curtailed MacArthur River and Key Lake in 2018 after a series of earlier partial reductions, by which point Cameco had 70% of its production shut in.

The company bought cheap spot material to fulfil its contracted sales portfolio and limited losses on the spread, even as care-and-maintenance costs at the idled operations ran $7 to $9 million a month. The spot material it was purchasing was largely Kazakh in origin, the same overproduction that had broken the price in the first place.

Kazakhstan’s behavior during this period carries a lesson the market has since absorbed. By producing at maximum volume through a collapsing price environment, Kazatomprom succeeded in becoming the world’s largest producer of a commodity it had completely devalued. The dividend it owed to the Kazakh state shrank in proportion to the price it had helped destroy.

The parallel to a Gulf petro-state that pumps aggressively into a falling oil market and discovers it has liquidated a national asset at the bottom of the cycle is fitting. Saudi Arabia took several such cycles to internalize the value of spare capacity and price management. Kazakhstan absorbed one catastrophic bear market and drew the correct conclusion faster. Its strategy today is indistinguishable from Cameco’s: contract first, produce after, treat in-ground uranium as inventory with value rather than as a cost to be minimized by maximizing throughput.

What every cycle in this history shares is the secondary supply wedge functioning as both cushion and amplifier. On the way down, secondary supply floods the market and extends the bear beyond what primary economics alone would produce. On the way up, the existence of that potential secondary supply caps the price signal that would otherwise incentivize new primary investment. For 6 decades the wedge was always there. That is no longer the case.

The Cold War stockpiles are gone, the megatons-to-megawatts program concluded in 2013, and utility inventories, while still meaningful, are working stocks rather than the vast strategic overhangs of earlier eras. The market emerging from this bear is the first in the modern history of the industry that has to be balanced by primary production alone.

The market entering this bull is the first in the modern history of the industry that has to be balanced by primary production alone, from producers who have learned, through painful experience, exactly what happens when they do not exercise discipline over their own output.

Next week, the second part of this analysis examines what that structural shift actually means in practice: how the two largest uranium producers in the world are now running explicit supply restraint simultaneously, why the geopolitical fracturing of the global uranium market is creating two separate procurement realities rather than one, and why the demand picture building against this constrained supply base is different in kind from any previous nuclear renaissance story.

The pattern that ended every prior uranium bull is structurally less available in this cycle. Understanding why requires understanding what has changed on both sides of the ledger at once.

Like this post or we’ll shut down the comment section until sentiment improves.

X-Energy was listed on the Nasdaq under the ticker “XE” on April 24th, raising $1.02 billion with shares surging 30.9% on debut to give the company a valuation of $11.9 billion.

The pitch is Dow Chemical as anchor customer, Amazon as both lead equity investor and holder of power purchase commitments for more than 5 GWe, and Centrica as a UK utility partner, all anchored to an 80 MWe pebble-bed high-temperature gas-cooled reactor.

X-Energy is a 916-person development-stage company promising rapid construction of high reliability reactors fuelled by proprietary TRISO fuel which is roughly 10 times more expensive to fabricate than conventional light water reactor fuel. For a deep dive on that topic, see Decouple’s interview with Michael Seely below.

It has no construction experience, has never operated nuclear reactors, and is yet to produce commercial fuel. Yet it is priced as though the hardest engineering problems in one of nuclear’s least mature reactor classes are already solved.

The current AI-driven moment of nuclear euphoria creates the conditions where this is possible. In the absence of a sustained tempo of nuclear construction in the West, narratives trump empiric reality and reactor concepts that are genuinely interesting science projects get priced as commercial products well before the performance record justifies it.

The embedded assumption in X-Energy’s $11.9 billion valuation is that American startup dynamism can compress what has been for China, the world’s most capable nuclear industrial state, a 50-year institutional journey into a single first-of-a-kind construction project.

In sectors where design, capital, and IP are dominant, that assumption is often correct. China’s pebble-bed program, however, illustrates why even “advanced nuclear” is not one of those industries.

Share

The German Origins and Chinese Technology Transfer of the HTGR

The pebble-bed concept originated with Rudolf Schulten at the Jülich Research Centre in Germany in the 1950s and 1960s. Germany built and operated the AVR and THTR-300 before political and economic pressure killed the HTGR program in the aftermath of Chernobyl.

China then absorbed 30 years of German engineering knowledge through formal technology transfer. This included a design review by Siemens/Interatom of the Chinese prototype HTGR, the HTR-10 in 1994. They have subsequently spent the next 3 decades building on it.

After 5 years of construction the 10MWt prototype reactor achieved criticality in 2000. Construction of the 80MWe commercial demonstration HTR-PM began in 2012, 38 years after basic research started at Tsinghua University.

It would be a mistake to treat those 38 years as a universal law of pebble bed development. For most of that period, especially prior to the 2000s, the program ran at the pace of a modestly funded national laboratory inside a much less capable economy with considerably more pressing industrial priorities than an experimental reactor concept. The China that initiated pebble bed research in the 1970s bore little resemblance to the nuclear industrial state that completed the HTR-PM in the 2010s.

A more concentrated, well-capitalized effort could plausibly move through the design and testing sequence faster, but the historical precedents for that kind of speed, the US Navy reactor programme and the AEC-era national laboratories were state-mobilized at a scale and institutional depth that a private startup has yet to replicate.

The Technology Readiness Level and Capacity Factor problem

The pebble-bed high-temperature gas-cooled reactor is a precise example of the gap between scientific promise of “advanced nuclear” and the commercial maturity of conventional water cooled reactors. Approximately 650 commercial PWRs and BWRs have operated since the 1950s, accumulating 18,000 reactor-years of operational experience.

Against that baseline, exactly 5 pebble-bed HTGR cores have ever operated: 2 pilot-scale research reactors, one German and one Chinese and 3 commercial demonstration-scale cores. Those were Germany’s short lived THTR-300 at 120 MWe (1983-1989) and a pair of 80MWe HTR-PM modules at Shidaowan, which reached commercial operation only in 2023.

Commercial demonstration scale pebble-bed HTGRs have thus accumulated only 12 reactor-years of operation to date. X-Energy’s customers are being asked to assume that its reactor technology will automatically share the 90%+ capacity factors that the light water reactor fleet has obtained through 18,000 reactor years of learning and optimization.

Dow and Amazon need high availability and the Xe-100’s central marketing claim is that continuous online refuelling delivers exactly that: no scheduled outages, dispatchable baseload and therefore high capacity factors from day one.

The best way for a lay person to make sense of the pebble-bed online refuelling concept is a giant gum ball machine. The pebble-bed HTGR continuously recirculates hundreds of thousands of fuel spheres through the reactor core, approximately 2000 per day, through a burnup measurement station, and back into the top of the core. This process must function with exceptional reliability as a continuous industrial operation rather than a periodic maintenance event.

As a comparison, the only other commercial reactor type that refuels continuously on a daily basis is the CANDU, which shuffles approximately 15 fuel bundles per day through a well-understood mechanical sequence refined across a fleet of reactors over 50 years. CANDU online refuelling still ranks among the more demanding operational challenges Canadian operators face.

The HTR-PM experience suggests this is harder to sustain than the design predicts. These refuelling difficulties may contribute to the fact that the Chinese plant entered commercial service derated 20% below its design thermal rating and reached only around 20% capacity factor in its first 2 years, advancing to around 50% by 2025.

Critically, China’s engineers entered the HTR-PM’s commercial demonstration phase carrying over 20 years of hard operational data from the HTR-10 research reactor. Problems like fuel handling anomalies, pebble flow irregularity, graphite dust management, helium leaks that manifested at 10 MWt research scale informed the commercial system’s design before a single pebble moved through it.

X-Energy has no equivalent operational foundation. Its fuel handling system has been modelled and tested at sub-scale, and it proposes to move directly from those results to a commercial build at Dow’s Seadrift facility.

The OPEX-CAPEX trade off

The 9 year construction record of the HTR-PM carries a specific warning for the X-Energy investment thesis. The business case for a low-pressure advanced reactor with expensive TRISO fuel, depends on offsetting high operating costs through faster, cheaper construction.

A pressurized water reactor must keep its coolant liquid at around 155 atmospheres, demanding thick-walled pressure vessels and elaborate emergency cooling systems that dominate capital costs and slow construction.

Despite its far lower power density and therefore relatively larger core, a helium-cooled pebble-bed reactor operates at lower pressure with fuel designed to retain fission products passively. The nuclear island should therefore in theory be simpler and faster to build.

Here a construction environment comparison matters as much as the reactor design. China’s construction sector since the 2010s is amongst the most capable nuclear building environments that have ever existed. Chinese builders have been simultaneously erecting conventional PWRs in under 6 years at 2 to 3 times the Western pace, drawing on deep supply chains, and a trained nuclear construction workforce. In that environment, the HTR-PM still took 9 years to build and a further 2 years to commission before entering commercial operation.

X-Energy will build in the contemporary American nuclear construction environment. That environment produced the only 2 AP1000 units completed in America in the 21st century at Vogtle, which came in at roughly 3 times the original budget and 7 years late, under cost-of-service regulation that insulated the project from market consequences throughout.

There is no equivalent rate-base backstop for Dow Chemical. The labour scarcity, the supply chain atrophy, and the absence of a standing nuclear construction workforce are structural features of the American sector, not problems that startup capital resolves. China’s HTR-PM first-of-a-kind timeline, built inside the contemporary world’s strongest nuclear construction infrastructure, is therefore a best case for X-Energy rather than a like-for-like comparison.

The bold claim made by X-Energy CEO Clay Sell in the present tense that “We make it easy to build nuclear power plants” will certainly be put to the test. Extraordinary claims require extraordinary evidence and that evidence has yet to be produced.

The Limits of American Startup Dynamism

X-Energy’s optimism relies on the assumption that American entrepreneurial dynamism will trump Chinese state development of advanced reactor technology.

This dynamism clearly works where iteration cycles are short, failure is recoverable, and the feedback loop between experiment and result is measured in months: shale drilling, rocketry and software are the cardinal examples.

SpaceX can destroy many prototypes on a single launch pad, absorb the data, and fly again within months. However, even an “advanced” nuclear reactor takes years to build with each reactor requiring its own seismically qualified civil works equivalent of a launch pad. In addition, its components must operate without failure for years and breaking things at the commercial reactor stage is not an option.

The iteration cycles, the failure intolerance, and the manufacturing precision of nuclear requirements impose longer timelines and slower iteration. This is what China’s HTGR program illustrates and no amount of venture or retail capital changes that.

But surely a peppy startup can move faster than a state run program in China?

The Western image of China as a cumbersome mass manufacturer of cheap goods rather than a technology leader is desperately out of date.

The Australian Strategic Policy Institute’s Critical Technology Tracker found China leading in just 3 of the technologies it tracked during 2003 to 2007. By 2024 that figure was 57 of 64, with nuclear explicitly among them. The country X-Energy is being benchmarked against, implicitly or otherwise, is not the China of 2003.

This is due to aggressive industrial policy and state capacity organized within a hybrid political economy. America’s nuclear reactor technology development dominance with the PWR/BWR similarly leaned heavily on robust state involvement to nurture the technology through its early development.

What China has built in pebble-bed HTGR technology over the last 5 decades is the capacity to cycle continuously from materials research through fabrication to operational deployment and back, accumulating compounding institutional learning at each pass.

X-Energy proposes to move from blueprints to commercial reactors at Dow’s Seadrift plant in under a decade, and according to Amazon’s commitment, envisions 5GW of capacity across the US by 2039.

X-Energy is a genuinely interesting technology company working on a compelling reactor concept. However, the entire history of pebble-bed HTGR development argues against the idea of leap-frogging the pilot and demonstration reactor stage and proceeding directly to a commercial build.

China ran the HTR-10 for over 2 decades before turning on its commercial scale demonstration reactor and is only now upsizing from its 2 module configuration to a fully commercial 6-pack, the HTR-PM 600 in the hopes of improving the small reactor’s challenging economics.

X-Energy has no equivalent to the HTR-10 or the HTR-PM. The comparison that ultimately matters is not how long China's program took in calendar years, but what China brought to its commercial demonstration that X-Energy does not: 2 decades of dedicated research reactor operational data, the world's most capable nuclear construction sector, and cost-of-service regulation backstopping the learning curve.

$11.9 billion is an immodest price for a firm that, by its own timeline, proposes to do something more ambitious than what China attempted, with less preparation and in a more difficult construction environment.

The American and European reactor construction booms of the 1970s and 1980s have run their course. Western utilities, saturated with capacity after decades of rapid industrial build-out, have stopped ordering reactors. Chernobyl has done its political damage.

The French Messmer programme has delivered 54 reactors in roughly 15 years, but Paris is now operating a mature fleet, and not expanding it. In the United States all reactor orders placed after 1973 have been canceled. The word most commonly attached to nuclear construction, on both continents, is “pause.”

Japan is an exception. It is the world’s second largest economy dependent on imports for 90% of its energy needs. Driven by energy security concerns it has developed a national target of 50% nuclear electricity by 2030. Japanese utilities have continued building at scale as the West has stood down.

TEPCO, the Tokyo Electric Power Company, operates large multi-unit sites at Fukushima Daiichi, Fukushima Daini, and the world’s largest nuclear generating site, Kashiwazaki-Kariwa. It has brought a systematic industrial discipline to the task that its Western counterparts have largely abandoned.

At Kashiwazaki-Kariwa, units 6 and 7 come online. They are ABWRs, 1350 MWs each and the world’s first Generation III reactors. The first of a kind (FOAK) unit is built in 38 months of nuclear construction and costs roughly a third less than unit 5, a seventh-of-its-kind standardized BWR5 Mark II, built at the same site by the same construction crews, essentially simultaneously.

This stellar result sets the benchmark for the EPR designers at Framatome and the AP1000 team at Westinghouse. Both programmes publicly aim for 36-month construction schedules. As Krellenstein observes, this prompts a reasonable inference: if the Japanese can build a FOAK Gen III reactor in 38 months with 1990s digital tools, why couldn’t the West?

The ABWR appeared to prove that the obstacles plaguing nuclear construction in the 1970s and 1980s were not intrinsic to the technology. They were interpreted primarily as problems of document management, design coordination and supply chain integration, precisely the problems that the information technology revolution of the 1990s was supposed to have solved.

Share

A Controlled Experiment

When Olkiluoto 3 began construction in 2005 as the first EPR, the Western nuclear industry had the same computers, the same CAD software, the same building information modelling tools, and the same AutoCAD and Primavera and Navisworks that Japan had used to build Kashiwazaki Kariwa 6 and 7.

Commodity laptops were faster than anything available when TEPCO planned its first ABWR. Adobe Acrobat was on every engineer’s desktop. Radio frequency identification (RFID) inventory management became standard practice in manufacturing supply chains across the developed world.

However, results in the West differed markedly from the Japanese experience. Finland’s Olkiluoto 3 took 18 years from construction start to grid connection. The Vogtle AP1000 units in Georgia, the only AP1000s completed in the United States, ran roughly 7 years late and billions over budget, bankrupting Westinghouse in the process.

The EPR at Flamanville in France, begun in 2007, achieved grid connection only in 2024. These projects were assembled after a long pause, dependent on teams, supply chains, and regulators lacking recent experience in delivering nuclear megaprojects.

Digitalization was necessary for Japan’s impressive achievement, but it was only one variable in the stack. Japanese nuclear construction continuity into the 1990s conserved a workforce with recent build experience, project managers who had closed out large units, and a supply chain still qualified and synchronized to nuclear specifications.

Kashiwazaki Kariwa 6 and 7 were successive iterations of an ongoing construction program. Japan demonstrated pre-construction discipline, an integrated supply chain, the utility as developer model, and above all spent more than a decade on pre-construction design and constructability work.

The 1978 Tokyo Meeting

The ABWR originated in a TEPCO feasibility study commissioned in 1978, 13 years before construction began at Kashiwazaki-Kariwa. TEPCO brought together every major boiling water reactor developer in the world: General Electric, Hitachi, Toshiba, the Italian firm Ansaldo, and the Swedish firm ASEA, which had independently developed BWR technology without a GE licence, deploying it at Forsmark, Ringhals, and Oskarshamn.

Such cooperation was deliberately radical for an industry organized around vendor competition. TEPCO was paying its vendors to make their existing designs obsolete in order to secure the next one.

The premise underlying their assignment was expressed in one question: if you could redesign the boiling water reactor from scratch, drawing on everything the global BWR fleet had learned about construction and operation, what would you change?

That meeting produced the conceptual design that eventually became the ABWR. Its underlying philosophy was simple: build what the utility could actually construct, validate, operate, and understand, using proven components and methodologies available at the time, and eliminate complexity wherever it did not buy proportionate safety or performance.

This is quite different from the philosophy that produced the AP1000, which emerged primarily from US government-funded research into advanced passive safety concepts. The AP1000 is revolutionary in its passive safety architecture. The evolutionary ABWR made design decisions driven primarily by the needs of a utility aiming to build at scale.

The Design Philosophy

The ABWR containment building is a squat cylinder of reinforced concrete, shorter than the turbine building on the same site. Height is both a key driver of cost and an enemy of seismic resilience. Japan’s geology demanded a short, stiff structure.

The free volume of the ABWR containment is roughly 1/5th that of an equivalently sized PWR, 450,000 cubic feet, compared to 2.5 to 3 million cubic feet in an equivalent pressurized water reactor. Most of this reduction in containment volume is intrinsic to boiling water reactor design: with the elimination of steam generators, pressurizer, hot legs, or cold legs. However, the elimination of external recirculation pumps reduced containment size further relative to other BWRs.

Perhaps the most revealing design choice of the ABWR is the reactor internal pump, or RIP. To understand why, it is worth recalling what came before.

Boiling water reactors benefit hugely from recirculation flow. In a BWR, water in the core is both coolant and neutron moderator. Steam bubbles, which are roughly 100 times less dense per unit volume than liquid water, weaken moderation and reduce reactor power. Controlling the void fraction in the core (reducing steam bubbles) is therefore a primary mechanism of power control.

The way to do this is to drive recirculation flow through the core: pushing water through the fuel at high velocity sweeps bubbles away and increases moderator density. An extraordinary illustration of the importance of recirculation flow is the fact that its flow rate dwarfs the feed water flow rate, that is the water returning to the reactor pressure vessel from the turbine condenser, by a factor of 5-7.

Early BWRs accomplished this with large external recirculation loops: pipes carrying reactor water out through the vessel wall, through pumps, and back in at the bottom. GE’s BWR3 through BWR6 product lines replaced those loops with jet pumps: 2 external recirculation pumps driving high-velocity jets into a Venturi throat, with surrounding water entrained by the Bernoulli effect, multiplying the flow without multiplying the large-diameter pipe penetrations. Most operating BWRs in the United States today possess this configuration.

The ABWR went further. Drawing on technology pioneered by ASEA on the Swedish fleet at Forsmark, designers eliminated external recirculation entirely, replacing it with 10 RIPs mounted on the bottom skirt of the reactor pressure vessel. The RIP motors sit outside the vessel; the impellers are on the inside, driven through the lower skirt. There are no recirculation loop pipe penetrations. The only fluid penetrations below the midpoint of the nuclear fuel are the RIP penetrations themselves, which are far smaller in diameter than the large recirculation piping they replaced.

The resultant safety gain is real: recirculation line breaks contributed significantly to large break loss-of-coolant accident probability in earlier BWR designs. The ABWR eliminates that category of event entirely. But more importantly, as far as construction is concerned, is the simplification of the reactor building layout.

Removing the recirculation loops removes a substantial volume of large-diameter, nuclear-quality piping from the containment building, along with all the associated weld inspections, seal packages, and QA documentation. Every weld eliminated is a weld that cannot introduce a defect, cannot require a repair, and cannot become an inspection item for the regulator.

Operational consequences are equally significant. Each of the 10 RIPs is individually controlled by a variable frequency drive. By increasing pump speed and raising moderator density uniformly across the core, the reactor can move from 60% to 100% power in roughly 2 minutes without moving a control rod.

Krellenstein notes that this ramp rate is comparable to a combined-cycle gas turbine. Because the power change spreads evenly across the core, it avoids the local “hot spots” in neutron activity that happen when control rods are pushed in from specific locations.

The late-cycle “xenon pit” problem, which normally limits a reactor’s ability to ramp power up and down near the end of its fuel cycle, is less restrictive in the Advanced Boiling Water Reactor. Its internal recirculation pumps increase moderation in a way that preserves enough reactivity margin to keep adjusting power smoothly right up until the final coast-down period.

Thus, in a single piece of hardware, RIPs constitute a safety improvement, a construction simplification, a maintenance reduction, and an operational flexibility gain. That combination is the physical manifestation of ABWR’s pragmatic design philosophy.

The Original Large Modular Reactor

The ABWR is a large modular reactor. That framing matters because modularity is currently marketed as a defining advantage of small modular reactor designs, as though off-site prefabrication become available only below a certain nameplate capacity. The ABWR demonstrated otherwise at 1,350 megawatts.

Every unit ever built, including the Lungmen plants in Taiwan, was constructed on a bargeable site, with a large crawler moving directly from the barge slip to place the entire bottom section of the containment as a single piece onto the foundation. The fabrication shops producing those modules were inside the same “Zaibatsu” vertically integrated conglomerates as the design organizations, so the integration between what was designed and what could actually be shipped and placed was well understood.

The modularization schema was deliberately flexible, which turned out to matter more than the degree of modularization on any given unit. The first ABWRs were more stick-built, with prefabrication increasing on the third and fourth plants as construction teams learned what the approach could and could not do. When the hydraulic control units, initially modularized as a single prefabricated room, proved harder to execute that way than anticipated, the teams pulled back and stick-built them instead.

In a Western programme governed by fixed contract milestones and adversarial relationships between the utility, the architect-engineer, and the vendors, that kind of adjustment generates claims and schedule impacts. At Kashiwazaki-Kariwa it generated a revised work package, because TEPCO owned the design and bore the consequences of getting it wrong, which meant it also had the authority to fix it without litigation.

Four Hours Sharpening, Two Hours Cutting

The ABWR’s extraordinary built time was largely due to pre-construction planning. 13 years of design development, regulatory engagement, supply chain preparation, and digital pre-construction work occurred prior to nuclear concrete being poured.

By the time Kashiwazaki-Kariwa 6 broke ground, the design was finalized. Every cable tray, every pipe penetration, every weld sequence, every construction work package had been resolved in the digital model. Worker congestion through the construction sequence had been reduced. Concrete pours had been scheduled to the seasons to meet weather requirements. The entire warehouse inventory was managed by RFID. QA travellers and field work packages were transmitted to Palm Pilots on site rather than managed through paper binders.

From rock inspection to commercial operation, every subsequent ABWR built in Japan was completed within roughly 40 to 50 months. The 4 Japanese ABWRs operational by 2006 were built at a staggering rate of approximately 28 megawatts per month of construction time, or roughly 1 MW per day. For sake of comparison Vogtle Unit 3 was only 9 MW/month and the Flamanville EPR 8 MW/month.

The standard aspiration of nuclear construction improvement is a learning curve story: the FOAK is hardest, and each subsequent unit gets cheaper and faster as the builder learns. The ABWR bucked that trend. Its FOAK was a third cheaper than the seventh of a kind predecessor it replaced. The learning happened in the context of an active build program and 13 years of construction-informed design work before ground was broken, not in the field afterward.

In contrast the AP1000 and EPR were not merely unfinished designs when they entered the field, they were attempting to compress the Japanese pre-construction process into the construction schedule itself. They paid for every unresolved interference and every incomplete work package in real time, with tradespeople standing idle and interest during construction sky rocketing.

The failure to export the ABWR

In 2003, Finland’s utility Fortum selected the EPR over the ABWR for Olkiluoto 3. The EPR slightly underbid the ABWR, and from the perspective of a utility making a 2003 decision, the choice was not obviously wrong.

Kashiwazaki-Kariwa 6 and 7 were already operating as reference plants, but the industry had not yet absorbed what its benefits were. The prevailing and seemingly reasonable assumption was that the industry could now expect Gen III construction performance in the 36-to-38-month range. Olkiluoto 3 was supposed to be the proof point for the EPR just as Kashiwazaki-Kariwa had been for the ABWR.

The Olkiluoto 3 result is now part of a sorry historical record: 14 years late and roughly 3 times the original budget.

Would these outcomes have been partially avoided if The ABWR had been built in Finland or the United States? The ABWR had genuine advantages that the EPR and AP1000 lacked at first pour. The Lungmen units, 2 ABWRs built in Taiwan, that although never allowed to operate for political reasons, were built to ASME codes and represented a direct analogue to a Western regulatory environment.

The ABWR was the first reactor ever certified under the NRC part 52 framework. As mentioned above, the modularization scheme was more flexible than the AP1000’s rigid modular architecture, which contributed significantly to field execution problems at Vogtle.

Toshiba’s withdrawal from the South Texas Project in 2018 represented a financial failure, not a technical one: unpromising market conditions and an inability to find investors after the shale gas revolution collapsed merchant power economics were the spoilers. Hitachi’s suspension of the Wylfa Newydd ABWR project in Wales in 2019, after it had completed the UK Generic Design Assessment in 2017 and spent roughly £2 billion on development, was due to a deficiency in the financing structure, and not the reactor design or its licensing record.

Does the US need competition in large reactor technologies?

As mentioned in the introduction to this essay the U.S. Department of Energy has begun exploring alternatives to the AP1000. To the DOE, Westinghouse and its majority owner, the private equity firm Brookfield Asset Management, are moving too slowly. Utilities, for their part, want cost-overrun insurance that the current federal financing framework does not provide, and Westinghouse’s 2017 bankruptcy from Vogtle cost overruns is not a distant memory for anyone being asked to sign a new contract. Reporting by Alex Kaufman in March 2026 describes active engagement with GE Vernova Hitachi Nuclear Energy on potential deployment of the Advanced Boiling Water Reactor, alongside separate outreach to South Korean counterparts regarding the APR-1400.

The case for competition has precedent within the US nuclear industry itself. In the 1990s and 2000s, steam generator replacement outages across the operating fleet initially required roughly 200 days. Over successive projects, that window was compressed to approximately less than 30 days.

The improvement came from repetition across a series of similar projects combined with direct competition between Bechtel and Steam Generator Technologies (SGT). Utilities could observe divergent performance in real time. Schedules could no longer be defended as fixed physical constraints. Each iteration forced contractors to explain delays in operational terms rather than attributing them to inherent complexity.

A similar dynamic is visible in current reactor construction programmes abroad. China maintains parallel large reactor supply chains, building Chinese AP1000 units alongside the Hualong One. This structure keeps cost and schedule performance legible across competing designs and organizations.

The technical basis for introducing competition in the United States already exists. The ABWR holds a completed design certification in the United States, with multiple reference units constructed and operated. The remaining constraint is institutional rather than technological.

Does GE Vernova have the capability to build both large and small modular reactors?

The most important constraint on an ABWR revival is organisational capacity. GE Vernova through GE Hitachi Nuclear Energy, which holds the ABWR certification, operates with roughly 1,400 to 1,500 employees globally and is already fully committed to first of a kind BWRX 300 deployments in Tennessee and Ontario.

Those programmes require sustained focus across licensing, supply chain formation, and site execution. A large reactor programme would introduce a parallel effort with distinct engineering, procurement, and regulatory demands.

This reflects a structural shift. The organization that developed the ABWR and built the BWR5 and BWR6 fleet operated at larger scale within an industry that was building continuously. Today’s entity is smaller, more focused, and aligned around small modular reactor deployment, with capital and strategy directed toward the BWRX 300 and a reduced physical footprint following the sale of Vallecitos, GE’s historic BWR development site in California.

Alex Kaufman reports that the company has limited interest in reactivating the ABWR. An organisation of this size is unlikely to be able to sustain two separate reactor programmes simultaneously. The case for competition in reactor procurement remains, but it depends on whether any developer has the scale and execution capacity to deliver it.

Focusing on the wrong problems

The Western nuclear debate is hyperfixated on novel reactor designs as the solution to its decades-long struggle to build nuclear on budget and on schedule. The hope seems to be that a technological breakthrough or a novel concept will create a demonstrably superior reactor that will reverse decline and restore competitiveness with China.

Modularity, passive safety, novel coolants, TRISO fuel: these grab the headlines and drive funding cycles. Reactor design is no doubt a part of the equation. Modularity done well reduces field labour. Passive safety systems can eliminate backup diesel generators, reduce concrete and steel use and simplify cabling.

The ABWR however, illustrates that in nuclear the most consequential design gains often come from seemingly mundane construction and operations informed improvements rather than novelty. Nuclear technology iterates slowly and mature light and heavy water designs have taken 40+ years to arrive.

History however is clear that design alone, however well-conceived, is a relatively small portion of the ingredients that determine whether a reactor programme succeeds. Light water reactors despite significant design differences between PWRs and BWRs for instance have all converged on similar ranges of capacity factors, economic performance and construction schedules. There has not been a clear winner.

As the ABWR experience illustrates, what seems to really matter to successful deployment is design completion prior to construction and delivery by a competent project development organization, operating inside an industrial ecosystem that is active and staffed with experienced people at every level from project managers to tradespeople.

The West is currently disadvantaged after decades of relative inactivity within a broader shift in political economy towards deindustrialization and financialization. Supply chains have atrophied and experienced professionals are mostly retired.

As a result the Western nuclear renaissance, when it comes, will likely be a crawl before it learns to run again. Early projects will disappoint when measured against Japan’s record from the 1990s or China’s contemporary construction timelines. The central challenge of revival will involve sustaining political and financial commitment in spite of those disappointments while maintaining a laser focus on the lessons the ABWR experience has taught us.

Listen to the full conversation with James Krellenstein on the Decouple podcast.

Ten RITM-200 reactors have been manufactured since 2012 for five Project 22220 icebreakers, with eight already operating on the Northern Sea Route aboard the icebreakers Arktika, Sibir, Ural, and Yakutia. The RITM-200S is the FNPP optimized variant

As previously discussed on Decouple, rather than emerging from a clean-sheet SMR program the RITM descends in an unbroken line from the first generation of Russian nuclear submarines through three further generations of icebreakers accumulating over 400 reactor years of operation.

The mine these reactors will serve is not a marginal resource play. The Peschanka field within the Baimskaya licence area contains Joint Ore Reserves Committee (JORC) standard resources of 9.9 million tonnes of copper and 16.6 million ounces of gold, with annual ore processing capacity projected at 70 million tonnes. There is no grid, gas pipeline, or hydroelectric resource within reach, and diesel at this scale and remoteness is not a serious economic option.

Why Not Gas Turbines?

The choice of nuclear over gas turbines was not purely a matter of Arctic physics. Russia is no stranger to building gas infrastructure north of sixty; however, its large-capacity gas turbine industrial base has never been genuinely domestic.

High-capacity units above 100 MW were sourced overwhelmingly from Siemens, GE, and Alstom through licensed technology, joint ventures, and direct imports. The vulnerability of that arrangement became visible well before 2022. Siemens halted deliveries to Russian state-controlled entities in 2017 after four of its turbines were illegally diverted to Crimea in breach of EU sanctions. Then, after February 2022, GE stopped all servicing of its Russian-installed base under expanded US sanctions prohibiting engineering services.

Russia’s first genuinely domestic high-capacity gas turbine, Rostec’s GTD-110M, reached its inaugural commercial installation only in October 2024; serial production capacity of four units per year is not expected until 2028. At the scale required by Baimskaya, no sanctions-proof gas turbine solution existed or was imminent when the power contracts were drawn up.

Nuclear offered complete vertical integration within the Russian state system, spanning reactor design, fuel fabrication, hull construction, and the refuelling logistics chain at Atomflot, with no real Western sanctions exposure. The higher upfront capital cost was the price of autarky, and given the trajectory of Western sanctions policy, it was the rational price to pay.

The FNPP hulls are Chinese, built by Wison Heavy Industries at its Nantong yard, selected through competitive tenders on cost and schedule grounds over Russian and South Korean alternatives. Domestic yards run simultaneously at capacity on the Project 22220 icebreaker program.

Unlike Wison’s Zhoushan entity, sanctioned by the US Treasury for its work on Novatek’s Arctic LNG 2 project, the nuclear hull contract was placed with Wison’s Nantong facility, a legally and operationally distinct yard within the same corporate family. The Nantong facility’s supply chain, financing, and labour pool sit outside the transactional surface which Western sanctions were designed to reach. Thus no dollar-clearing bank, Western component supplier, or counterparty with secondary sanctions exposure are required.

Western Maritime Nuclear Myopia

Rather than carefully studying the Russians, the undisputed leaders in maritime SMRs, Western commentators tend to track NuScale’s regulatory milestones, Core Power’s fundraising announcements, or the latest molten chloride fast reactor rendering.

NuScale and Canadian firm Prodigy Clean Energy have proposed housing up to 12 NuScale Power Modules on a single marine facility, a concept that scales to 924 MWe on paper. Yet as former Decouple guest Aleksey Resvoi has pointed out, this proposal must contend with NuScale’s 20-meter tall “sky scraper” integral power module and containment vessel, which presents obvious stability problems aboard any hull experiencing real sea conditions.

An analogy which illustrates the difference in integral steam generator design philosophy between NuScale and the RITM is the maintenance gap between the American M1 Abrams and the Russian T-72. The Abrams is a genuine engineering achievement but runs on a 1,500-horsepower gas turbine which cannot be meaningfully repaired outside a depot, and whose filters clog in mud.

The T-72 runs on a diesel engine that a regimental workshop can overhaul. Soviet designers drew the line between crew-level and specialist maintenance differently, and drew it much further forward.

NuScale’s helical coil steam generators are thermodynamically clever, achieving compact geometry and high heat transfer surface area inside the integral vessel. They are, however, challenging to access and maintain.

The RITM uses replaceable steam generator cassettes with an explicitly designed 20-year service life, shorter than the reactor pressure vessel’s 60-year rating by intent. This accepts a modest thermodynamic penalty in exchange for cassettes that can be more easily extracted and replaced during scheduled maintenance. Major overhauls are conducted at Atomflot, Rosatom’s purpose-built nuclear fleet-servicing base in Murmansk, the only non-naval facility in the world with the infrastructure to handle operating maritime reactors.

Core Power has gone further into novelty territory here, partnering with TerraPower and HD Hyundai to pursue TerraPower’s Molten Chloride Fast Reactor (MCFR) for marine application. The MCFR proposes to use liquid chloride salt as both fuel and coolant, to operate at fast neutron spectrum, and to run at higher temperatures than conventional reactors. It has progressed from a non-nuclear integrated-effects test, running pumped-salt operations in a laboratory in Everett, Washington, towards a non-power critical experiment at Idaho National Laboratory targeted for approximately 2030.

Molten salt chemistries are corrosive, require exotic materials, and have never been operated continuously at commercial scale in any environment, let alone one subject to rolling, pitching, and wave slamming.

Pragmatism Wins the Day

The engineers who built the RITM understood that every design decision had to survive contact with the Arctic. They were not optimizing for thermodynamic elegance, waste minimization, or fuel cycle closure.

Russia kept iterating on light water because light water is proven. They now have 400 battle tested reactor-years of maritime operating history having harnessed the same design bureaus, test loops, materials laboratories, and institutional memory within a state-directed vertically integrated industrial system.

The West, while nostalgically clinging to narratives of nuclear energy dominance, keeps betting that start up culture and conceptual novelty will leapfrog its competitors’ steady investment, continuous deployment and iteration of proven technologies.

The RITM embodies everything Western policymakers say they want from an SMR: modularity, serial production, flexible deployment across marine, barge, and land-based applications, and a demonstrated fuel cycle.

With eight reactors already operating in Arctic conditions and 8 more under construction or fitting out, Baimskaya is the proof-of-concept that makes Rosatom’s FNPP export pitch credible in a way unmatched by any current Western maritime nuclear proposal.

The Floating Nuclear Power Plant Revolution may not be televised, but it will be Russian.

Imagine a technology that could take the most feared byproduct of nuclear power and transform it into fuel: burning the long-lived actinides that make spent fuel a supposed million-year liability down to fission products that decay to background radiation within a few centuries.

All of this while extracting enough energy from the 95,000 metric tons of nuclear waste sitting in pools and dry casks across the United States to power the entire country for 150 years.

That technology has been demonstrated. At Idaho National Laboratory, the Experimental Breeder Reactor II (EBR-2) ran for 30 years on metallic fuel cycled through an attached electrochemical reprocessing facility, closing the loop in exactly the way the vision describes: spent fuel in, fresh fuel out, waste reduced to a centuries-scale rather than geological-scale problem. The physics and engineering has been proven at pilot research scale. It has however failed to scale.

What did scale commercially looks considerably different. La Hague, the world’s largest reprocessing facility, has operated for the better part of a century. France built it not out of a romantic attachment to the closed fuel cycle but out a perceived energy security necessity: Almost no domestic fossil fuels, and, in the postwar decades, uranium supply chains that looked dangerously thin.

French reprocessing works. It has extended the country’s uranium utilization, deferred the need to mine fresh uranium by putting depleted enrichment tails and recovered plutonium back to work as MOX fuel, and given Paris a degree of fuel cycle independence that looks increasingly valuable as the theme of energy security once more rises to prominence with the Iran war and the closure of the Strait of Hormuz.

What French reprocessing does not do is meaningfully reduce the volume of material destined for deep geological disposal. But the La Hague program is not EBR-2. It is a pragmatic industrial process operating within the constraints of thermal reactor physics, and those constraints are severe. The gap between what France actually achieves and what the closed cycle promises is the central story of nuclear reprocessing, and it begins with understanding what is actually sitting in those spent fuel pools.

Subscribe now

What Spent Fuel Actually Is

When uranium fuel completes its working life in a reactor, it leaves behind a material of remarkable complexity. A fresh fuel assembly loaded into a pressurized water reactor contains uranium enriched to roughly 5% U-235, the fissile isotope that sustains the chain reaction, with the remaining 95% being U-238, which does not fission under thermal neutron bombardment but plays its own important role. Over the course of 18-24 months in the reactor core, several things happen simultaneously.

The uranium-235 is progressively consumed, splitting into fission products: a zoo of mid-weight atoms, many of them intensely radioactive, which accumulate in the fuel and increasingly poison the neutron economy in the reactor. Separately, some of the uranium-238 absorbs neutrons and transmutes into heavier elements. The most significant of these is plutonium-239, which is fissile and begins contributing meaningfully to the reactor’s power output. By the time a fuel assembly is discharged, roughly half of the energy being generated comes from this in-situ produced plutonium rather than from the original uranium-235.

What emerges from the reactor is therefore not simply depleted uranium. It is approximately 95% uranium, mostly uranium-238 with perhaps 1% uranium-235 remaining, around 1% plutonium spread across several isotopes of varying usefulness, and 3 to 4% fission products.

The fission products are what make spent fuel so radiologically hostile: short-lived but intensely radioactive isotopes generating substantial decay heat and penetrating gamma radiation. The barrier to reloading the spent fuel is primarily that the fission products have made the neutronic environment unworkable and the physical handling of the fuel intolerable. Reprocessing is, at its core, the project of separating the still-useful heavy metals from this contaminating matrix.

Share

The Manhattan Project’s Civil Legacy

The motivation for developing that separation capability did not originate in civilian power generation. Plutonium-239’s properties as a weapons material, chiefly its relatively modest critical mass compared to highly enriched uranium and its producibility from abundant uranium-238, made its extraction a military priority from the earliest days of the American atomic program.

The Hanford site in Washington State was separating plutonium for the Fat Man bomb before a single civilian reactor had generated a watt of commercial power. PUREX, the plutonium-uranium redox extraction process that remains the dominant reprocessing technology today, was a product of that program.

The civilian rationale came later, and it was genuinely compelling given what was known at the time. Postwar estimates of global uranium reserves suggested severe scarcity. If nuclear power was to provide a substantial fraction of the world’s energy, a once-through fuel cycle would exhaust available uranium within decades.

Reprocessing and breeding, in which fast neutron reactors would produce more fissile material than they consumed, appeared not merely attractive but necessary. Scientists at the Chicago Metallurgical Laboratory, working through the physics with the tools available to them, were not wrong to conclude that closed fuel cycles were the only path to long-term nuclear viability.

What they could not know was that uranium would prove far more abundant than early surveys suggested. As exploration intensified, reserves expanded dramatically, and the economics of fresh fuel remained stubbornly favorable relative to the cost of recycling.

When Jimmy Carter banned commercial reprocessing in the United States in 1977, framing it as a nonproliferation measure, the ban was already pointing at a technology that market forces had largely sidelined.

The commercial record made this plain. West Valley in New York, the only U.S. reprocessing plant that ever operated, ran from 1966 to 1972 before shutting down; the estimated cost of upgrades needed to meet newer regulatory requirements rose above $600 million, making restart uneconomic.

Barnwell in South Carolina was partially constructed but never operated: the 1977 policy reversal interrupted construction, and a subsequent GAO review found little commercial interest in completing it given a weak nuclear outlook, unresolved federal policy, and high costs.

The GE Morris plant in Illinois never operated as a reprocessing facility at all; cold testing revealed the technology was more complex than anticipated, the plant would have required extensive modification, and it became a spent-fuel storage site instead. When Ronald Reagan reversed Carter's ban in 1981, the commercial response was revealing: nothing happened. No reprocessing industry emerged. The economics had already delivered the verdict that a rowdy policy debate had obscured.

Inside La Hague

France is the exception that proves the rule, and La Hague is where that exception takes physical form. The largest reprocessing facility in the world, handling around 1700 metric tons of spent fuel per year (equivalent to roughly 17-20 full core loads) at peak operation, La Hague has been running in various forms since the late 1950s. Understanding what it actually does, step by step, is the only way to understand what commercial reprocessing can and cannot accomplish.

Spent fuel arrives in massive shielded transport casks, each containing perhaps a dozen fuel assemblies and weighing enough to require specialized rail or road transport. The assemblies must sit in cooling pools for months before processing begins, allowing the most intensely radioactive short-lived fission products to decay and reducing the thermal load to manageable levels. This waiting period involves a tradeoff that will become important later: the longer the fuel cools, the more certain isotopes decay into daughter products that complicate downstream handling.

When processing begins, the fuel assemblies are fed into a guillotine chopper: a hydraulically powered blade that slices the entire assembly, zircaloy cladding, ceramic fuel pellets and all, into segments roughly three centimeters long. This is the moment when cladding that has maintained a hermetic seal through 4.5 years of neutron bombardment releases in a single instant the fission gases that have been accumulating inside under pressure. Krypton, xenon, iodine and other volatile species must be captured and filtered rather than vented to atmosphere, and they constitute the first gaseous waste stream the process generates.

The chopped fuel segments then drop into boiling nitric acid. The acid dissolves the uranium dioxide ceramic and releases the heavy metals into solution while leaving the zircaloy cladding hulls behind as a solid residue. Those hulls, too contaminated by neutron activation and fission product migration to recycle economically, are compacted and stored as an intermediate-level waste stream.

What remains after the acid dissolution is a solution containing uranium, plutonium, fission products, and minor actinides all dissolved together. This solution is then contacted with tributyl phosphate dissolved in a kerosene like hydrocarbon solvent. Here is where the chemistry gets elegant: uranium and plutonium, unlike the fission products, preferentially migrate into the organic phase. By manipulating the oxidation states of uranium and plutonium relative to each other, exploiting the fact that they behave differently under different redox conditions, the process separates them into distinct streams with around 99.5 percent recovery efficiency. The fission products, stripped of the heavy metals they were mixed with, remain in the aqueous phase with almost no cross contamination.

Those fission products are then concentrated, calcined, and incorporated into borosilicate glass, poured into steel cylinders and stored at La Hague. If the fuel came from a foreign utility, France is contractually and legally obligated to return these cylinders to the country of origin, which has created its own geopolitical complications over the years with Japan and Germany. The uranium stream and plutonium stream depart the process separately, each carrying their own challenges.

What the Process Actually Reduces

It is worth pausing on a claim the reprocessing narrative tends to treat as settled: that PUREX substantially reduces the volume of waste requiring deep geological disposal.

The standard figure, repeated by the World Nuclear Association and the IAEA alike, is an 80 to 85 percent reduction in high-level waste volume compared to a once-through cycle. That number is technically defensible in one narrow sense. It compares the volume of vitrified fission product glass canisters against the volume of intact spent fuel assemblies, and since uranium makes up roughly 94 percent of spent fuel by mass, removing it does dramatically shrink the pile of material formally classified as high-level waste. But the comparison omits something important.

In a once-through fuel cycle, the depleted uranium tails from enrichment are a low-level waste problem at most. They are mildly radioactive, carry no fission products, and do not require geological disposal.

When France takes that material, mixes it with separated plutonium to fabricate it into MOX fuel, and irradiates it in a reactor, the depleted uranium that would never have required a deep repository has been converted into spent MOX fuel that under current conventions does.

In addition spent MOX fuel is several times more radioactive than equivalent spent LEU fuel, and its decay heat declines more slowly, meaning it occupies more effective repository space per assembly than the fuel it replaced.

Reprocessing generates multiple additional waste streams that intact direct disposal does not: compacted zircaloy hulls and hardware from fuel disassembly, low- and intermediate-level liquid wastes from the PUREX chemistry, contaminated resins and solvents, and solidified residues from effluent treatment. Each of these streams requires its own handling, conditioning, and disposal pathway.

The total volume of material requiring regulated disposal across all these categories exceeds what would have resulted from simply burying the original spent fuel assemblies whole. Reprocessing does not reduce the waste problem so much as it disaggregates and redistributes it across a more complex and costly management chain.

Standard PUREX also leaves the minor actinides, americium, curium, and neptunium, in the fission product waste stream rather than recovering them. These are the isotopes with very long half lives responsible for most of the radioactivity in the hundred-thousand-year timeframe that regulators believe make deep geological repositories a civilizational-scale engineering necessity.

What reprocessing in a thermal reactor cycle does avoid is one fresh mining cycle and one batch of enrichment tails: a real but narrow benefit, achieved at significantly greater expense and effort.

The Isotopic Trap

The plutonium stream that emerges from PUREX is not weapons-grade material, and understanding why illuminates one of the central constraints on the entire reprocessing enterprise.

Plutonium-239, the isotope produced when uranium-238 absorbs a single neutron, is fissile and desirable both for weapons and for fuel. But uranium-238 does not stop absorbing neutrons after producing plutonium-239. Successive captures produce plutonium-240, which is not thermally fissile and constitutes a neutronic burden; plutonium-241, which is fissile; and plutonium-242, which is not. The longer fuel remains in a thermal reactor, accumulating burn-up, the more the plutonium inventory shifts toward these higher isotopes.

A reactor optimized for weapons-grade plutonium production would discharge fuel after only a few weeks, with burn-up measured in single-digit gigawatt-days per tonne. A commercial power reactor runs its fuel to fifty or sixty gigawatt-days per tonne, producing plutonium that is substantially degraded from a fissile quality standpoint.

This isotopic degradation compounds with each reprocessing cycle. In a thermal reactor, plutonium-239 and plutonium-241 are consumed preferentially during irradiation while plutonium-240 and plutonium-242 accumulate.

After one pass through a MOX fabrication and irradiation cycle, the recovered plutonium contains a significantly higher fraction of non-fissile even numbered non-fissile isotopes. After a second pass, the material is approaching the point of limited utility in a thermal neutron spectrum. The twice-through-then-out cycle that France actually operates is not a policy choice made for lack of ambition; it reflects the physics of what thermal reactors do to plutonium.

The uranium stream carries its own complications. Reprocessed uranium (RepU) is not natural uranium. It contains uranium-236, a neutron absorber that reduces reactivity and uranium-232 which relatively quickly decays through a chain that produces thallium-208, an energetic gamma emitter that makes the material progressively more difficult to handle as it ages. To put it coarsely, instead of aging into a fine wine, reprocessed uranium deteriorates into mouldy grapes. Fabricating it into fresh fuel requires higher enrichment levels and handling radioactive feedstock on production lines designed for essentially inert natural uranium powder, demanding remote or shielded operations where direct human contact was previously routine.

At La Hague, the recovered uranium stockpile has grown for decades. France stores it in the hope of eventual use but has not found an economically or practically compelling route to deploying it at scale. It is the program’s largest stranded asset.

MOX: What the Plutonium Actually Becomes

The separated plutonium does have a destination. Mixed oxide fuel, combining recovered plutonium with depleted uranium tails from the enrichment process, is loaded into 15-20 of France’s fifty-four operating reactors. This is the practical expression of the entire PUREX enterprise, and it functions but with constraints that reveal the limits of the technology in a thermal reactor context.

Fresh LEU fuel contains roughly five percent uranium-235 as its fissile component, with uranium-238 making up the remainder. MOX fuel substitutes plutonium for that fissile loading, with the fissile plutonium fraction typically running between 4 and 8% depending on the isotopic quality of the recovered material and the desired burnup target. The depleted uranium carrier, with its negligible uranium-235 content, is essentially inert from a fission standpoint. From an energy density perspective, the two fuels are broadly comparable in a reactor core

.

The neutronic differences are real, however. Plutonium fission produces a shorter delayed neutron lifetime than uranium-235 fission. Delayed neutrons, the tiny fraction of neutrons emitted seconds after fission rather than instantaneously, are what make thermal reactors controllable on human timescales. A reactor core with a higher plutonium fraction responds more rapidly to reactivity perturbations, tightening the margins that operators and control systems manage.

France limits MOX loading to around 30% of any given core not because higher fractions are physically impossible, but because they would require operators trained and equipped for a reactor that behaves differently from every simulator they have used.

The AP1000 and EPR were designed with full MOX capability in mind, incorporating the additional control authority that a 100 percent MOX core demands, but that capability has not been commercially exercised.

A CANDU solution to the RepU problem? Natural Uranium Equivalent (NUE)

The accumulating stockpile of reprocessed uranium (RepU) at La Hague poses a difficult challenge. CANDU reactors, which run on natural uranium and use heavy water as both moderator and coolant, operate with such a thermally efficient neutron economy that they can sustain criticality on fuel enriched well below the threshold required for light water reactors. RepU, with its residual uranium-235 content of around one percent, sits above the natural uranium threshold and looks, at first glance, like a natural feedstock for a heavy water reactor. The advanced fuel CANDU reactor concept, pursued through a joint venture between Candu Energy Inc. and China National Nuclear Corporation attracted enough political weight to bring the Chinese Premier Li Keqiang and Justin Trudeau to the same signing table.

The concept has genuine logic. Canada possesses the reactor technology capable of using the material. China, which is expanding its commercial reprocessing capability will soon possess RepU in quantities that would otherwise sit indefinitely in storage. This RepU could be fashioned into Natural Uranium Equivalent fuel (NUE) by downblending it slightly with depleted uranium from enrichment tails to be used in CANDU reactors. Both parties saw the potential benefits. What the demonstration program revealed was how much difficulty lies between an appealing concept and a functioning fuel fabrication process.

The small-scale fabrication of the 24 NUE bundles manufactured for the demonstration irradiation was more difficult than foreseen. Specifically selected batches of depleted uranium were needed to match the batches of recycled uranium to achieve an equivalent weight percent uranium-235 to fabricate the pellets.

There were problems with dry blending the depleted uranium and recycled uranium for sintering the pellets, because the grain sizes of the uranium dioxide powder had very large variations. To overcome this issue, the material had to be dissolved in acid and put through a conversion process back to uranium dioxide powder, essentially reintroducing a wet chemistry step into a fuel fabrication process whose simplicity was part of the CANDU value proposition. The challenges with large-scale production were not resolved. NUE has not been used further in the Qinshan reactors.

The reasons reach beyond fabrication difficulty. A CANDU fueling machine loads individual bundles into pressure tubes that run horizontally through the reactor, with fresh and spent bundles moving through each channel during power operation.

A core loaded with RepU bundles of variable enrichment introduces neutronic asymmetries that the standard operating model was not designed to manage. The uranium-232 decay chain imposes a time pressure on using the material before it becomes too radioactive for conventional handling. Lastly the reprocessed uranium’s isotopic composition, variable across different source batches and different cooling histories, resists the kind of precise characterization that stable reactor operation demands.

Each obstacle individually might be surmountable; collectively, they have proven sufficient to stall the program.

The Fast Reactor Promise and Pyroprocessing

Everything described so far takes place in the thermal reactor world, where neutrons are moderated to low energies and the isotopic limitations of recycled plutonium become progressively more binding. The genuine closure of the nuclear fuel cycle, the version where ninety-five thousand metric tons of American spent fuel might theoretically power the country for a century and a half, requires a different neutron spectrum entirely.

Fast reactors, in which unmoderated neutrons sustain the chain reaction, can fission isotopes that thermal reactors cannot. Plutonium-240, the accumulating burden of every thermal reprocessing cycle, fissions readily under fast neutron bombardment. So do the minor actinides: americium, curium, neptunium, that constitute the long-lived hazard in spent fuel.

A fast reactor with a closed fuel cycle does not merely extend uranium utilization; it progressively burns down the transuranic inventory that makes geological disposal a deep-time engineering problem. The fission products that remain are genuinely short-lived on geological timescales, decaying to background levels within a few centuries rather than hundreds of thousands of years.

The reprocessing technology best suited to this vision is pyroprocessing rather than PUREX. Instead of dissolving spent fuel in nitric acid and manipulating liquid-liquid extraction chemistry, pyroprocessing dissolves metallic fuel in a molten chloride salt and applies an electrical potential across electrodes.

Heavy metals migrate electrochemically, leaving fission products behind. The process does not separate plutonium from uranium or other actinides for that matter. They all migrate together, which is simultaneously a proliferation advantage and an operational feature for fast reactor fuel cycles that want to keep them mixed. EBR-2 at Idaho National Laboratory ran for decades on metallic fuel with an attached pyroprocessing facility, demonstrating the integrated concept at a small research scale.

The gap between that demonstration and commercial deployment encompasses the entire history of fast reactor development. Despite heroic efforts by the USA, Japan, France, India, Russia and China, no fast reactor has achieved the reliability and economics of the mature light water reactor fleet. In fact the only fast reactors currently operating at all are two units in Russia, one in China and one in India, the PFBR, which achieved its 1st criticality on April 6th of this year after more than 20 years of construction.

The fuel required for pyroprocessing, metallic uranium-plutonium-zirconium alloy, is used in no commercial power plant anywhere in the world including those Russian, Chinese and Indian fast reactors all of which use MOX.

Pyroprocessing has been demonstrated in batches measured in kilograms, not the hundreds of tonnes per year that a meaningful contribution to spent fuel disposition would require. The economic structure of building new pyroprocessing capacity is significantly impacted by capital cost recovery. If built the early decades of operation will impose a heavy fixed-cost burden regardless of throughput. La Hague can offer relatively competitive reprocessing because its capital costs are long since written down; a new pyroprocessing facility would need to recover its full construction cost against a fuel market priced on the marginal economics of existing enrichment capacity. Oklo’s peak valuation of $21 billion reflects the market’s appetite for the narrative more than its assessment of the near-term industrial reality.

The Permanent Condition

America’s 95,000 metric tons of spent fuel await disposal. The physics and processes that make them potentially valuable is proven and has not changed. Uranium-238, transmuted to plutonium and cycled through fast reactors with pyroprocessing, could in principle extend the energy yield from existing fuel stocks by factors that dwarf what PUREX and MOX achieve in practice.

What the history of reprocessing demonstrates however, is that the distance between physical principle and economic reality is vast. As with many compelling solutions to vexatious problems the pilot program exists. Scaling and making a process commercially viable remain the core challenge determined in this case by isotopic decay chains, tight fabrication tolerances, material accounting protocols, and high capital costs amortized over small batches of fuel for reactors that have failed to become commercially viable.

France has run the world’s most ambitious civilian reprocessing program for the better part of a century and extended its uranium utilization by roughly 10% at the fleet level. This is a meaningful achievement but far from being transformative.

Nuclear energy lives permanently in this tension. It must bridge the gap between compelling narratives and the grinding cost and difficulty of the engineering.

Reprocessing has arrived at the destination the physics and economics permitted, which turned out to be considerably short of where the dream was pointing.

The complexity of conventional oil sits at the importer’s end in multi-billion dollar refineries which take years to build.

LNG is almost the opposite: its complexity sits at the exporter’s end, locked inside exquisitely engineered liquefaction infrastructure. On the importer’s side warming the liquid back to its gaseous state requires little more than running seawater over a heat exchanger.

The Germans demonstrated this at breakneck speed during the energy crisis of 2022. They built 5 floating storage and regasification units, essentially modified LNG tankers moored offshore. The fastest was stood up in under seven months.

Once regasified the pressure is managed, a pipeline is run to a combined cycle gas turbine, and voila you have electricity. Those power plants are assembled in 24 to 36 months from factory-built components with very low construction risk. They achieve close to 60 percent thermal efficiency, making them the most thermodynamically elegant large-scale power plants ever constructed.

The complexity on the production side is a different matter entirely. Each liquefaction train at Ras Laffan represents roughly $7 billion in capital investment. The process reduces its volume by roughly 600 times through cryogenic cooling, getting it down to the minus 162 Celsius threshold where it occupies a manageable volume for ocean transport.

Subscribe now

This is the asymmetry that explains why so many countries made what seemed like rational energy security bets on LNG. The capital burden on the importing side was manageable and the long term contracts relatively affordable. Supply during an era of open sea lanes and reliable contracts, looked stable.

Unlike crude oil, where pricing is anchored to deep, globally traded benchmarks and a highly liquid spot market, LNG remains structurally tied to long-term bilateral contracts necessary to cover the high capital cost of liquefaction facilities.

Today roughly 60% of global LNG volumes are committed under such contracts, leaving a relatively thin pool of uncontracted supply. Most cargoes are pre-allocated, so a country reliant on Qatari volumes cannot readily switch to Australian or U.S. supply at short notice. In a global supply crunch these buyers are pushed into the spot market, where competition for a relatively thin pool of uncommitted cargoes drives prices sharply higher.

Even as the U.S. brings new liquefaction capacity online, LNG supply remains slow to adjust. New trains are costly and can take close to a decade from initial planning to first cargo, leaving limited spare capacity to absorb sudden disruptions.

The 2022 European gas crisis offered a deceptive lesson. Germany solved it by building receiving infrastructure fast, which made LNG look resilient. That was however an importing problem, not a production problem. Global liquefaction capacity was intact and US production was surging. The receiving end of LNG bends to political will, tight timelines and emergency permitting. Destroyed liquefaction infrastructure does not.

In the year or so since it became clear that the second Trump administration and Republican Congress were intent on dismantling the Inflation Reduction Act (IRA), a particular intellectual formation has hardened among climate-focused commentators in the West.

The argument runs as follows: China has identified that the industrial commodities of the 21st century as solar panels, batteries, electric vehicles, and the rare earth magnets inside wind turbines and EV drivetrains. By pouring state capital into manufacturing all of these at an overwhelming scale, China has positioned itself as the dominant industrial power of the coming era.

The United States, by contrast, has chosen fossil fuels and is becoming a museum of obsolete technologies, to use Bill McKibben’s phrase. China, in this reading, is not just winning an industrial competition. It is saving the climate.

Historian Nils Gilman, writing in Foreign Policy, envisions an “eco-ideological Cold War” pitting a potential Sino-European green entente against a US-led petro-bloc. Noah Smith writes that China is “quietly saving the world.” James Meadway declares that China has “obliterated” the argument that nothing can be done on climate because of China.

The electrostate thesis, as it has become known, holds that China is reorganizing its entire economy around electricity, increasingly sourced from wind and solar, and that this transformation is both a play for industrial supremacy and an inadvertent path to global decarbonization.

There is enough truth in this picture to make it influential. China installed 315 GW of new solar capacity in 2025 bringing its cumulative installed PV capacity to 1.2 TW nearly the entire installed capacity of the US grid. China’s solar manufacturing dominance is extraordinary. Its battery supply chain is, by most measures, unassailable on any near-term horizon. Its electric vehicle industry has crossed an inflection point that now looks permanent. These facts are not in dispute, and the cost reductions China has driven in solar and batteries have been genuinely transformative.

The problem is that the thesis is constructed almost entirely from what is visible at the output layer, from the finished solar panel, the EV on the road, the battery pack. Follow the supply chain upstream, ask what industrial system produced those things and what powered that system, and a very different picture emerges.

There is a Ford Motor Company promotional film from the mid-twentieth century that traces the full manufacture of an automobile from the first blast of explosives in an iron ore mine, through smelting, rolling, stamping, and final assembly. Its purpose was to celebrate American industrial integration. It functions equally well today as a corrective to how Western observers think about Chinese manufactured goods. We see the finished product and construct a story about the civilization that made it. We do not however ask what was burned to produce it.

The Ballast Rock

China’s own government is clearer about its energy system than its Western admirers tend to be. Official policy statements and domestic media describe coal as the “ballast rock” of China’s energy supply, providing cheap electricity, industrial heat, and chemical feedstocks for the bulk of the industrial economy. Coal generated 59 percent of Chinese electricity in 2024. China consumes 56 percent of all coal used globally. Vanishingly little coal capacity has been fully retired, only 2.5 GW in 2024.

The construction pipeline makes the trajectory plain. The desert industrial parks in Xinjiang tell the same story: more than half a dozen coal generation units, several of them commissioned since 2022, supplying aluminum reduction cells, metal silicon furnaces, and the ultra-pure polysilicon plants that feed China’s solar manufacturing supply chain.

Shaanxi’s coal to chemicals industry, centered on the city of Yulin, continues to expand with government backing: the National Development and Reform Commission (NDRC) has classified coal-derived chemicals as a ‘new productive force’ in the same policy breath as solar panels and electric vehicles.

Here’s a fact for China’s climate cheerleaders to mull over. As Javier Blas has pointed out, if China’s coal to chemicals industry was a country it would rank as the world’s third largest consumer of coal, trumped only by the rest of the Chinese coal sector and India’s national consumption, but ahead of the major coal users like the US, Japan, Indonesia and Turkey.

The claim that China is becoming a green electrostate sits alongside an operational coal fleet of 1300 GW with 261 GW of coal capacity either under construction or already permitted. While some are ultra-supercritical coal plants which now serve as flexible backup and load-smoothing assets for variable renewables, this fleet is not a legacy sector approaching retirement. Many are fresh investments in the sectors the electrostate narrative holds up as proof of China’s clean industrial future.

China will not retreat from coal for ideological reasons, as the European bloc has done with both coal and nuclear. Chinese planners have reasons to retire older, less efficient plants and reduce coal combustion near urban centres, but there is no structural incentive to remove a working power source from a system still growing at roughly 7 percent per year in electricity consumption.

Aluminum as a Diagnostic

An insightful case study for what is actually happening in China’s industrial economy is aluminum, sometimes referred to as “congealed electricity.”

Making a ton of primary aluminum requires roughly 13,000 kilowatt-hours (kWh) of electricity. To put that in terms that register: producing one kilogram of aluminum consumes about as much electricity as running an average North American house for twelve hours, while recycling a kilogram of steel in an electric arc furnace (EAF) takes roughly half an hour of equivalent draw.

The aluminum industry has always chased the cheapest available power. It built smelters next to the Bonneville Power Administration and the Tennessee Valley Authority (TVA) in the United States, next to dedicated hydroelectric dams in Iceland and Norway, and in Quebec, where legacy hydro produces electricity at costs no thermal plant approaches. American aluminum, which has shrunk to one sixth of its peak production of 4.6 million tonnes/yr in 1980, used to receive half its power from federal or state-owned utilities.

China now produces roughly 60 percent of the world’s primary aluminum, about 45 million tons per year, against a domestic capacity cap set by the NDRC to prevent further oversupply.

The electrostate narrative around Chinese aluminum runs as follows: production is migrating from the coal heartlands of Shandong province toward Inner Mongolia and Xinjiang, where renewable energy development has accelerated, and toward the hydropower-rich southwestern provinces of Yunnan and Sichuan.

The implication is that one of the world’s most electricity-intensive industries is being progressively decarbonized by the same renewable buildout reshaping China’s broader grid.

The timeline punctures this immediately. The majority of new aluminum capacity in Inner Mongolia and Xinjiang was built before 2020, before large-scale wind and solar deployment in northern China, and before the central government introduced any renewable energy purchasing requirements for the sector.

These smelters relocated to the northwest because coal was cheap there, and local governments were offering captive coal power at rates that made the economics irresistible.

The renewable energy purchasing requirement, mandating that smelters source a portion of their electricity from clean generation, was introduced only in 2024, and currently sits at around 30 percent for Inner Mongolia and Xinjiang.

Even on paper, even counting purchased renewable energy certificates rather than electrons physically flowing from wind turbines to reduction cells, these smelters remain predominantly coal-powered.

And paper is doing considerable work in that sentence. Upward of 80 percent of aluminum smelters in Inner Mongolia, and practically all of those in Xinjiang, have on-site captive coal power plants. There is nothing transitional about that infrastructure. The renewable purchasing requirement’s 30% cap is itself an admission that the economics of running these smelters on wind and solar do not close.

The Yunnan story is more favourable to the green narrative. Southern China’s hydropower provinces have genuine surplus generation capacity, and the Hongqiao Group, operating its aluminum business under the Weiqiao brand, has been aggressively moving capacity to Yunnan to capture cheap hydro before the NDRC production cap forecloses further expansion.

About 15 percent of China’s aluminum capacity now sits there. This represents genuinely lower-carbon production, but it is opportunistic: companies racing to claim cheap electrons while they still can, not a planned industrial transition toward renewable power.

Taking the full picture, roughly 25 percent of China’s aluminum production is hydropower-sourced. The other 75% runs on coal.

The Un-Electrification of Magnesium

If aluminum complicates the electrostate thesis, magnesium inverts it entirely. Before the late 1990s, most of the world’s primary magnesium was produced electrolytically, through a process broadly similar to aluminum smelting, using seawater-derived magnesium salts reduced through electrolysis. Production was spread across the United States, Norway, France, and Canada. The process was electricity-intensive but coal-free.

China industrialized magnesium production through a completely different route. The Pidgeon process, a pyrometallurgical method using carbon-based reduction of magnesium-bearing ores in small batch-operated furnaces, requires almost no electricity.

The coal chemical heartland of Shaanxi province became the global capital of magnesium production not because it had clean power or innovative electrolysis technology, but because it had coal to burn and workers to tend furnaces, and the economics of attaching a small magnesium shop to an existing coal chemical complex made the product nearly free at the margin.

The result was the complete elimination of Western electrolytic magnesium production. Today, over 95 percent of the world’s primary magnesium, a market of roughly one million tons per year, is produced in China through the Pidgeon process, almost entirely as a byproduct of coal chemistry.

Magnesium matters beyond its own market. It is a critical input for aluminum alloys and lightweight automotive and aerospace structures, which means it sits directly in the supply chain of the very EV revolution the electrostate narrative celebrates.

However, the electrostate, in this corner of the periodic table, ran in reverse. Magnesium production was deliberately un-electrified by China to achieve cost dominance and the climate movement has nothing useful to say about it because it contradicts the entire narrative.

The Electrotech Stack and Its Foundation

Zoom out from aluminum and magnesium to the full set of supply chains the electrostate argument relies on, and a structural pattern becomes visible that the narrative consistently elides.

The five supply chains at the core of the electrostate thesis, solar cells, lithium-ion batteries, rare earth permanent magnets, semiconductor-grade silicon, and the metals that go into EV structures, all share characteristics that are more explanatory than their clean end uses.

In each, China accounts for 75-94% of global production. The consumer-facing product sits downstream of an upstream processing step that is not modular, not easily replicated, and not captured by the language of the tech economy.

The electricity intensity of each of these upstream steps is extraordinary: solar-grade polysilicon requires five times as much electricity per unit mass as aluminum. Battery graphite matches aluminum’s energy consumption. Synthetic graphite, which accounts for 97 percent of global graphite anode production and is produced almost entirely in China, was built up in coal-intensive provinces precisely because the energy-intensive reduction process was cheapest where coal was cheapest.

China’s dominance in these industries was built between roughly 2005 and 2020, during the last great wave of coal capacity construction, when cheap and reliable thermal power allowed Chinese firms to accumulate the process knowledge, engineering expertise, supply chain integration, and economies of scale that now make these industries effectively uninvestable for outside competitors.

A cohort of chemical engineers who graduated from Chinese universities between 1980 and 1995, and spent their careers building polysilicon reduction plants, battery cathode factories, and rare earth processing facilities, is the actual foundation of Chinese industrial dominance in the electrotech stack.

The solar panels and wind turbines are the visible surface of a structure whose load-bearing elements were built in coal smoke, and the timeline matters: the dominance preceded the renewables buildout, rather than resulting from it.

What Chinese Energy Policy Actually Is

The electrostate framing imposes an ideological coherence on Chinese energy policy that the evidence does not support. China’s strategy is better understood as systematic, unsentimental diversification across every available energy vector simultaneously, driven by the specific vulnerability of importing roughly 12 million barrels of oil per day, much of it transiting the Strait of Hormuz.

China electrifies not because it has identified electricity as the industrial commodity of a new era, but because liquid hydrocarbons represent its central strategic exposure and domestic electricity generation is the domain where it has resource options.

China doubled its domestic natural gas production between 2015 and 2025 and will likely surpass Iran this year to become the world’s third-largest gas producer.

It has 200 GW of gas-fired generation capacity, roughly four times its nuclear fleet by installed capacity. Renewables are one strand of a deliberately plural strategy, and the electrostate thesis mistakes that strand for the organizing principle of the whole.

The think tank Ember Energy has described China’s coal flexibility as evidence of innovation, praising the retooling of coal plants as ideal partners for solar and wind. Ember’s framing is a good example of how the green electrostate thesis works in practice: China’s clean energy activities are foregrounded, and everything that complicates the picture is reframed as further evidence of sophistication.

Forget Climate Change, Can America Catch up in Aluminium?

The United States finds itself in an uncomfortable position. With aluminum tariffs applied to treaty allies, a domestic smelter fleet whose newest facility entered service in 1980, and primary aluminum production of 680,000 tons per year against consumption many times that figure the situation is grim.

The tariff logic is not entirely without merit: rebuilding capacity in strategically important metals is a legitimate national security objective, and China’s practice of building new coal-fired aluminum capacity in Indonesia and Vietnam as joint ventures, once the NDRC domestic cap was reached, demonstrates that the competitive pressure will not relent.

But the mechanism is badly matched to the problem. The average US aluminum smelter consumes 12 to 18 % more electricity per ton than its newest Chinese or Canadian competitors, a penalty reflecting decades of deferred investment. Power costs for US smelters exceed $60 per megawatt-hour (MWh); Canadian smelters pay $20 to $30.

The Aluminum Association has calculated that even a retrofit of an existing US smelter requires a power supply contract below $40/MWh to be financially viable. Tariffs move the commodity price. They do not move the electricity price, and the electricity problem is about to intensify, because a new million-ton smelter requiring 1.5 gigawatts of grid interconnection now competes for queue position with electricity hungry data centers operated by Microsoft, Google, and Amazon.

The structurally honest answer to the question of how the West rebuilds aluminum competitiveness might involve Alaska hydropower development and coordinated allied industrial policy with Canada, Europe, and South America, rather than the transactional trade war currently being waged against precisely those partners. Quebec’s legacy hydro system, producing electricity at costs no thermal plant can match, is a strategic industrial asset on the US border. Tariffing the aluminum it produces is an unusual way to secure an essential commodity.

What the Argument Requires You to Ignore

China emitting 25 % of global greenhouse gases is not a detail that can be resolved by pointing to its solar manufacturing output. The industries driving those emissions achieved Chinese dominance through a coal-powered learning curve that has not been reversed. The rate of emissions reduction in China’s industrial sector is slowing, not accelerating, as the low-hanging fruit of textiles electrification and cement demand decline gets exhausted.

If the goal is Western industrial and military competitiveness, the relevant question is what it would actually take to replicate or contest the system built between 2005 and 2020 by engineers who are now the most experienced in the world at what they do, working in an industrial ecosystem that was capitalized by cheap coal.

The answer is not flattering to any currently available Western policy agenda, which may partly explain the appeal of narratives that avoid the actual industrial mechanics of the competition.

The tariff instinct draws on a genuine historical memory: nineteenth century American industrial policy, the infant industry protectionism of the Hamilton Report and the high tariff era that followed, did help build domestic manufacturing capacity, and did so at a moment when the United States was catching up to British industrial dominance. The analogy is not absurd on its face. But that era involved building the underlying industrial infrastructure from scratch in a country with abundant coal, cheap land, and a rapidly expanding domestic market.

The current situation involves trying to rebuild capacity that was deliberately offshored over four decades, in an energy market where the cheapest available electrons are in Canada and Norway, against a competitor that accumulated irreplaceable process knowledge during a specific historical window that has now closed.

Nineteenth century tariff logic assumed you could build the thing if you protected the market long enough. The aluminum problem is not primarily a market problem. It is a power price problem, a supply chain knowledge problem, and a capital commitment problem, none of which a tariff schedule resolves.

On the climate front believing that China will take the world through the looking glass to net zero by 2050 requires believing that a country whose government describes coal as its energy ballast rock and which hosts captive coal power plants physically attached by conveyor belt to its most advanced manufacturing facilities, will undergo a transformation so complete and so rapid that it overrides the physical and economic inertia of all of that invested infrastructure.

The fact that this prediction finds a receptive audience in the Western climate movement is telling. It is what happens when people confronting what they believe to be an existential threat exhaust the available policy options.

The desperate need for a solution that produces a powerful appetite for evidence that someone, somewhere, is solving the problem at scale. China’s solar deployment numbers are impressive enough to feed that appetite. Everything upstream of the solar panel is easy to ignore, especially if looking at it means that the narrative collapses.

China’s clean energy manufacturing is real, consequential, and has permanently changed the economics of solar and batteries for the rest of the world. The industrial system that produced it is still, predominantly, powered by coal, and the Chinese government has told us so, repeatedly, in its own policy documents.

This essay accompanies a Decouple conversation with Seaver Wang, Climate and Energy Director at the Breakthrough Institute and co-author of “Greenwashing with Chinese Characteristics,” published in the Breakthrough Journal. If you find this work valuable, consider supporting Decouple by pledging on Substack or making a tax-deductible donation through our fiscal sponsor at Givebutter.

The Puzzle the Market Is Posing

Doomberg is back on Decouple. As one of the sharpest numerically-grounded voices writing about energy markets, he opened our recent conversation with a puzzle: given everything that has happened, the numbers simply should not look like this.

Venezuela’s president was arrested and extradited to New York, the United States imposed an embargo on Cuba, and a hot war with Iran has left the Strait of Hormuz closed and tankers sunk. Iranian missiles destroyed two liquefied natural gas (LNG) trains at Qatar’s Ras Laffan industrial complex, taking roughly 17 percent of Qatari production capacity offline for an estimated five years, while Ukraine continued its campaign against Russian oil tankers and refineries.

Had you presented that full scenario to any energy analyst six months ago and asked them to price oil, almost nobody would have said $98 West Texas Intermediate (WTI) and $110 Brent. Most would have taken the over on $150 without a second thought.

The relative calm in price action requires explanation. Oil’s share of global primary energy has fallen from around 50 percent at the time of the 1973 Arab oil embargo to roughly 35 percent today, meaning a given volume disruption transmits less shock into the broader economy than it once did.

The actual supply disruption running through the Strait of Hormuz is currently around eight to ten million barrels per day rather than the theoretical maximum of twenty, and China alone holds 1.5 billion barrels of strategic reserve, enough to absorb five million barrels per day of that shortfall for 300 days if deployed at scale.

None of this means the crisis is trivial. The pain is real and unevenly distributed. The question it poses, especially for Europe, is whether the shock is finally severe enough to force a reckoning with thirty years of decisions that left the continent structurally exposed in ways that were at least partially avoidable.

What the 1970s Actually Built in Europe

The European response to the OPEC embargo was an impressive mobilization that moved decisively and pragmatically taking advantage of the unique conditions available throughout the bloc.

France

France moved decisively launching the The Messmer Plan in direct response to the oil shock. It remains the fastest large-scale nuclear buildout in history. France had little domestic oil, but it had everything else the program required: a large corps of state-trained engineers produced by the Grandes Écoles, heavy industrial capacity rebuilt under the postwar dirigiste economic model, and a nationalized utility in Électricité de France (EDF) already accustomed to executing at the direction of the state rather than waiting on market incentives.

Its political class drew the logical conclusion: electrify aggressively around a domestic nuclear base, using a standardized reactor designs that a purpose-built supply chain could replicate at pace. Space heating, water heating, rail and significant portions of industrial process heat were shifted onto the grid as the fleet came online, deliberately substituting domestic electrons for imported hydrocarbons across as much of the economy as the technology of the era allowed.

The result over roughly twenty years was 54 operating reactors, an electricity system generating around 75 percent of its output from nuclear, net electricity exports to neighbours who had made different choices, and as an unintentional side effect a per-capita carbon footprint in the power sector that remains among the lowest in the developed world.

Britain Norway and the North Sea

Britain and Norway answered through a different channel with the same underlying instinct. The North Sea had yielded promising geological discoveries before the embargo, but the post-shock period transformed offshore petroleum into a central national project on both sides of the median line.

The technical obstacles were formidable: deepwater fields in rough northern seas, novel platform engineering, supply chain development for equipment categories that had barely existed before.

Both governments pressed through, operating under a governing assumption that sounds almost exotic in Brussels today, namely that a country should develop strategically important resources under its own waters when those resources can materially improve national resilience.

Within roughly a decade Britain had moved from significant oil importer to net exporter. Combined British and Norwegian output contributed materially to the global oil supply glut that emerged through the 1980s. The glut that always follows a supply shock once investment finally responds to price signals was this time in part European.

German Gas Diplomacy

West Germany’s answer came through diplomacy and welded steel. The first Soviet gas began flowing west through the Brotherhood pipeline in 1968, predating Ostpolitik’s formal articulation but embodying its central intuition: that a Russia earning deutschmarks from gas exports had a material interest in stable relations, and that pipelines, unlike armies, are hard to redeploy.

The 1973 embargo deepened that conviction. If imported Middle Eastern oil was vulnerable to political interruption, the answer was to secure more pipeline gas from a supplier with whom West Germany had a structured commercial relationship and a shared interest in its continuation.

The Urengoy-Pomary-Uzhgorod pipeline, completed in 1984 and running from the vast Urengoy gas field in western Siberia to the West German border, was a direct product of that post-shock logic, and it was built over fierce American objection.

The Reagan administration attempted to block European firms from supplying equipment and technology for its construction, arguing that the pipeline would create dangerous dependency. Bonn, Paris, and London essentially told Washington to mind its own business, and the pipeline was completed anyway.

What the gas bought Germany was an industrial cost structure that its European competitors could not match. German chemical producers, steel mills, glassmakers, and ceramics manufacturers built their economics around feedstock and energy costs that were simply unavailable to producers in Britain, France, or the United States.

BASF’s Ludwigshafen complex, the largest integrated chemical site in the world, was in part a monument to cheap Russian gas. Germany became the economic engine of the European Union partly on the strength of that bargain, running large industrial surpluses while its neighbors ran deficits, with cheap energy as one of the structural advantages that made German manufacturing so difficult to compete against.

The results of this collective mobilization were visible by the mid-1990s in a set of numbers worth holding in mind. Europe consumed around 32 EJ of hydrocarbons annually and produced around 15 EJ domestically. Roughly 8 EJ of Russian oil and pipeline gas, operating under long-term contracts with a generally reliable delivery history, functioned as quasi-domestic supply despite transiting through the iron curtain. Effective external import dependence was around 30 percent of consumption.

The French nuclear fleet anchored cheap, firm electricity across much of the continent. North Sea output was near its peak. Groningen, the vast Dutch gas field, underpinned Northwest European supply. Europe had a healthy energetic basis. A continent that had been genuinely vulnerable to a Middle Eastern supply cartel in 1973 had, through two decades of serious state-directed and private sector investment, rebuilt the physical foundations of its own security.

Share

The Pyramid Inverted

That position was dismantled through choices made by a political class that had become unmoored from the physical foundations of its own prosperity.

Maslow’s hierarchy was designed as a model of individual motivation, but it maps onto collective political behaviour with uncomfortable precision. A society that has secured abundant energy, industrial production, and physical safety creates the conditions in which political attention can migrate upward into questions of identity, moral positioning, and self-actualization.

The pathology arrives when those governing from the upper floors lose the ability to perceive the importance of the foundations beneath them, and begin making decisions that erode that base in the service of ideological commitments the foundations had made affordable.

By the mid-1990s, Europe had secured its ground floor thoroughly enough that its political class could afford to stop thinking about it. The first Conference of the Parties (COP) climate meeting opened in Berlin in 1995, the same year that marked Europe’s high-water point of energy self-sufficiency. That coincidence is not the whole explanation for what followed, but it is too precise to ignore.

The Decadence of Nuclear Phaseouts

No decision illustrates this dynamic more clearly than the closure of operating nuclear plants. The reactors that came online between the mid-1970s and late 1980s were not marginal facilities at the end of their useful lives. They were industrial cathedrals: vast, complex, irreplaceable concentrations of engineered knowledge representing decades of accumulated operating experience, trained workforces, regulatory culture, and supply chain depth, with low fuel costs, zero carbon emissions, and capacity factors that no other thermal generator could match.

They were shut down because opposition to nuclear power had become a marker of political identity in societies wealthy and secure enough to prioritize symbolic concerns over material ones. The antinuclear movement was a signal of ecological conscience and distrust of technocratic state power that resonated precisely because the lights had stayed on for so long that nobody could imagine them going out.

Germany shut its last three reactors in April 2023, in the middle of an acute energy crisis triggered by the loss of Russian gas, having spent the preceding decade replacing firm nuclear output with a combination of intermittent wind, coal, and imported gas.

The decision was not revisited when the crisis arrived, because doing so would have required the governing coalition to admit that the Green politics it had elevated to the centre of its identity had materially harmed the people it governed. Political identity proved stickier than physical reality, even once the electricity bills surged.

The scale of what was lost cannot be recovered quickly, and that is the point that cuts deepest against any optimism about Europe’s near-term capacity to respond. France still has a nuclear fleet, but it is the aging inheritance of a state that no longer exists in the form that built it.

The engineering workforce that commissioned those 54 reactors has largely retired. The specialized welding trades, the project management systems, the regulatory culture oriented around construction rather than obstruction, all of it was allowed to thin during the decades when new nuclear construction seemed unnecessary and the people demanding it were dismissed as relics of an earlier, less enlightened politics.

When France finally attempted to demonstrate that it could still build, the result was Flamanville 3: a single EPR reactor whose construction began in 2007, encountered cascading quality control failures and design changes, and achieved grid connection only in late 2024 after costs escalated to roughly 13 billion euros against an original estimate of 3.3 billion, a measure of how much institutional knowledge had been lost, and how expensive it is to rediscover what your predecessors once knew how to do.

Europe’s Hydrocarbon Dysmorphia

Today the continent consumes roughly 38 EJ against only 6 EJ produced domestically, with Russian supply now largely severed and Russian LNG under a 2027 phase-out deadline. This represents an extraordinary inversion of insecure hydrocarbon import dependency form 30% in 1995 to 80% in 2026.

The gap between what Europe consumes and what it can produce has widened from something manageable to something genuinely precarious.

The political economy behind this is not simple ignorance. It reflects the triumph of a specific set of assumptions that dominated European governance from roughly the mid-1990s onward. Neoliberalism and Ricardian comparative advantage, applied as governing doctrine held that offshoring industry and supply chains to wherever production was cheapest was rational efficiency rather than strategic vulnerability.

If someone else could supply energy more cheaply than Europe could produce it domestically, the logic ran, then importing was the optimal outcome, and if that someone else also absorbed the climate sin of extraction, so much the better. Europe could signal ecological virtue by de-industrializing while guzzling hydrocarbons produced out of sight and out of mind, congratulating itself on declining domestic emissions that had not so much been eliminated as outsourced.

The post-Cold War liberal international order would constrain the behaviour of exporters through norms, contracts, and multilateral institutions. The physical substrate, the reactors and rigs and pipelines, could be allowed to thin because the market would supply what was needed, and the moral ledger would look clean in the process.

Europe’s oil and gas consumption peaked at roughly 45 EJ around 2005 and has declined to around 38 EJ today, a reduction of around 15%. The European Environment Agency attributes this to a combination of efficiency improvements, warmer winters, higher post-2022 energy prices, renewables deployment, and structural shifts toward less energy-intensive industries, which is a diplomatic formulation for a process that deserves a blunter name: de-industrialisation.

Steel mills, chemical plants, ceramics manufacturers, and glassmakers that once anchored regional economies across the Ruhr, the Po Valley, and the Midlands have contracted or closed, taking their energy demand with them. The continent did not so much reduce its energy consumption as export the industries that consumed it, along with the employment, the tax base, the engineering knowledge, and the supply chain depth they represented.

Each factory that relocated to cheaper jurisdictions removed another supporting beam from the base of Maslow’s pyramid, trading the material foundations of working-class prosperity for lower headline energy consumption figures that Brussels could present as evidence of a successful transition. It was, in the most literal sense, progress measured by what had been lost.

What this framework systematically failed to account for was the difference between efficiency in a stable trading environment and resilience under supply stress. A just-in-time supply chain optimized for cost is the opposite of what you want when the supply is interrupted.

A grid optimized for cheap intermittent renewables is not the same system as one built around firm dispatchable capacity. The assumptions encoded in comparative advantage theory describe a world of frictionless markets and cooperative suppliers, which is occasionally the world that exists and is frequently not.

What Europe Could Still Do, and Why It Has Not Done It

Doomberg made compelling counter-arguments to my EU doomerism. He pointed out that California has roughly as much oil and gas geology as Texas. It produces a fraction of what Texas does because the legal, fiscal, and regulatory environment makes development effectively impossible.

The same circular logic operates in Europe. If exploration is prohibited, companies do not invest to assess and prove reserves. When reserves remain unproven, policymakers point to the low reserve figures as evidence that the resource is not there. The prohibition creates the condition it claims to be responding to.

Groningen, the Netherlands’ giant gas field, could be restarted. The decision to close it was driven by induced seismicity causing structural damage to buildings in Groningen province, a genuine local problem that nonetheless removed a strategically important asset from the European supply picture at precisely the wrong historical moment.

The Norwegian portion of the North Sea is still producing more than three times as much oil as the British side, and while Norway holds a larger share of the basin's reserves, the gap in output reflects policy as much as geology: Norway offers tax incentives for new development and drilled 33 exploration wells in 2025 while the UK spudded none.

Shale-bearing geology exists in Poland, northern England, France, and Germany. A 78 percent marginal tax rate on exploration and production in Britain, combined with a legal prohibition on hydraulic fracturing, does not describe a country that has concluded its geology is inadequate but rather a country that has politically foreclosed on its options.

The honest assessment is that Europe has real options it has chosen not to exercise, and that the human capital and industrial ecosystem required to exercise them quickly has been allowed to thin. A European shale revolution is perhaps not implausible on geological grounds, but it faces a 20-year American head start in drilling technology, a near-complete absence of the service industry infrastructure that makes rapid field development possible, and a regulatory and social environment that has been actively hostile to hydraulic fracturing.

Groningen could be restarted but not without political cost and not without the engineering assessment of what sustainable production levels might look like after years of inactivity.

New nuclear construction could be authorized but the supply chains and workforce needed to execute it at the pace the 1970s demonstrated are not reconstituted by political commitment alone.

What remains to be seen is whether Europe will experience the current crisis as severe enough to unlock the political will to change the framework. The fiscal terms, the regulatory culture, the legal prohibitions, the institutional preferences that have governed European energy policy for three decades are after all choices rather than facts of nature.

The question the crisis poses is whether the pain has finally reached the threshold at which those choices get revisited, or whether the current political class will manage this disruption as they did the 2022 energy crisis, through emergency imports and demand destruction with a return to the same framework when the immediate pressure eases.

China’s Different Theory of the State

The contrast that makes European failure most legible is not another European country that made better choices. It is China, whose energy policy over the same period reflects a fundamentally different answer to the question of what a state is for.

China’s 2024 primary energy figures from the Statistical Review of World Energy are worth reading slowly. Coal supplied 58 percent of primary energy, almost entirely from domestic mines. Nuclear, hydro, and renewables together contributed roughly 12 percent. Oil and gas accounted for the remaining 30 percent, of which about a quarter of the oil and 57 percent of the gas came from domestic production.

The effective Chinese import dependence on total primary energy is around 15 percent. Europe’s equivalent exposure is 80 percent on hydrocarbons alone, against a backdrop of reduced firm electricity generation following nuclear closures.

Beyond the headline numbers, China accumulated roughly an additional million barrels per day of crude oil throughout 2025 and stored it in strategic reserves. It built coal stockpiles. It invested in coal-to-chemicals facilities that convert coal through gasification into synthesis gas, the carbon monoxide and hydrogen mixture from which Fischer-Tropsch catalysis can produce diesel, aviation fuel, ammonia, and essentially any hydrocarbon compound a modern economy requires.

Germany developed Fischer-Tropsch synthesis in the 1930s and 1940s under blockade conditions to fuel the Luftwaffe when oil imports were cut off. South Africa built a large coal-to-liquids industry during the apartheid sanctions era. No country builds capital-intensive, carbon-intensive, economically irrational coal-to-liquids capacity for fun.

Countries build it when they are explicitly planning for fuel denial or war. China’s coal-to-chemicals program makes national security sense and no other kind of sense, which is a direct statement of how Beijing’s planners assess their strategic environment.

China also moved to dominate the processing stages of commodity supply chains where it could not dominate the upstream resource. It produces roughly five million barrels of crude oil per day domestically and imports around eleven million, but it has built refining capacity of nearly nineteen million barrels per day, a nameplate figure that dwarfs its own consumption and positions it as a swing supplier of refined products to the rest of the world.

When China banned refined fuel exports within days of the conflict escalating, Australia found itself in difficulty almost immediately. Australia had allowed most of its domestic refining capacity to close over the preceding decade on the assumption that cheaper Asian-processed products would always be available, above all Chinese jet fuel, which had come to supply roughly a third of the country’s aviation needs.

The vulnerability ran from Australian dependence on Chinese refined products through Chinese dependence on Middle Eastern crude to Middle Eastern supply now disrupted: a chain of three links, each individually visible, none apparently prompting precautionary action.

The pattern is consistent across rare earth processing, battery manufacturing, solar panel production, and refining: identify the processing stage where dominance can be created rather than relying on geological endowment and use that position as a buffer against upstream interruption.

This is preparation for a world where the Strait of Malacca can be contested, the first island chain can be used as a coercive instrument and the assumptions of the liberal international trading order do not hold under pressure. Europe read the post-Cold War decades as evidence that those calculations had become obsolete. China saw them as an interval of relative quiet before patterns that have governed great power competition for centuries reasserted themselves.

Can Europe Rebuild its Energy Foundations?

Doomberg holds as axiom that supply shocks reliably produce the investment responses that eventually create gluts. The 1973 embargo produced the North Sea, the Messmer Plan, the IEA strategic reserve system, and eventually a global oil surplus by the mid-1980s. The 2008 financial crisis and $147 nominal crude produced the capital desperation that drove the U.S. shale revolution to commercial viability. Something analogous will emerge from the current crisis, though its precise shape is not yet clear and the degree to which Europe contributes is questionable.

The optimistic reading is that the current disruption finally forces decisions that the 2022 crisis did not. The options are real: Groningen could be restarted, North Sea acreage could attract development capital under different fiscal terms, and Europe’s commitment to new nuclear construction, if sustained, will in theory produce operating capacity, though the timeline will be measured in decades rather than years.

The political conditions that produced the Energiewende, that justified closing operating nuclear plants in the middle of an energy crisis, were themselves the product of Europe’s masterful response to the 1970s energy shock.

The question now is whether Europe can turn its focus to the material basis of its well-being and restore its energy foundation.

The SMR craze reflects a fundamental category error and combined cycle gas turbine envy. It tries to map the attributes of a CCGT plant onto nuclear, where the underlying cost structure and physical infrastructure is fundamentally different.

A CCGT plant is essentially a jet engine bolted to a heat recovery steam generator and a smaller steam turbine. The critical point is where the complexity sits. The gas turbine, which is the expensive and technically demanding component, is built in a factory, hot functionally tested and shipped to site as a finished machine.

Construction on site is largely installation, foundations, piping, electrical connection, using conventional materials and repeatable processes. That architecture shifts risk into manufacturing and compresses timelines. Rather than building the hardest part on site you are simply installing it in 24-36 months.

Nuclear does not behave this way. It can incorporate modular components, but the NSSS is only 25-40% of cost. The dominant cost drivers sit elsewhere. Civil works, excavation, basemat, containment, seismic qualification, remain site specific and labour intensive. Nuclear grade quality assurance, documentation, and inspection add another layer of fixed overhead. Safety systems with redundancy and independence are function driven, not size driven, so they do not shrink proportionally with output.

The nuclear steam supply system is not analogous to the gas turbine in a CCGT. It is not a fully integrated, factory proven machine that arrives ready to run. The plant comes together on site, under regulatory oversight, with integration, testing and certification happening during construction & commissioning.

This is why economies of scale are so strong in nuclear. Many of the costs do not scale linearly with power. When you reduce reactor size, you reduce output and revenue, while a large share of the cost base remains. Studies show that smaller reactors actually increase the relative share of on site construction because the civil works do not shrink in proportion to capacity.

The SMR thesis assumes nuclear can transition from a project to a product, capturing the modular, factory built economics of gas plants. The constraint is that the parts of nuclear that dominate cost remain stubbornly project based.

The envy is understandble. CCGTs are extraordinarily compelling. They are marvels of thermally efficiency, capital light, fast to deploy and supported by a global supply chain of standardized components. They are the most successful large scale power plants of the past decades.

It is only natural that nuclear developers would look at that model and attempt to emulate it but in so doing they are committing a grave category error, an error that is setting the western nuclear industry up for decade(s) of disappointment.

Some SMRs will get built but they will not replicate the CCGT promise. They will be mini versions of large reactors with mini revenues to pay off the significant inherent costs of nuclear power.

Joe St. Julian arrived at AtkinsRéalis four years ago as an outsider. He had spent two decades at Bechtel, most of it in the American federal nuclear sector, running cleanup operations at the K-25 gaseous diffusion facility in Oak Ridge, then serving as project manager on the Hanford Waste Treatment Plant, the $25 billion vitrification facility built to immobilize the liquid plutonium production waste accumulated between 1945 and 1987. Before that he had built large combined-cycle gas plants in Michigan for the Shaw Group. He is, in other words, a construction and project controls man who spent his career in the most complicated industrial environments the American government could devise. He did not grow up in CANDU.

That background turns out to matter for what he had to say about it. When St. Julian arrived in Canada and started doing what he calls “due diligence” on the reactor technology he had just been hired to sell, he was surprised. “I was actually shocked,” he told me. “I thought it was an incredibly innovative technology.”

American nuclear culture produced four original equipment manufacturers whose light water reactor designs account, directly or indirectly, for the vast majority of the world’s operating fleet. Canada went a different direction entirely. No enrichment dependency, no large forged pressure vessel, a reactor that runs on natural uranium and refuels online without shutdown. Thirty-five units built across six countries with a further 18 CANDU inspired pressurized heavy water reactors built by India. Almost nobody outside those countries can describe how it works.

St. Julian’s diagnosis of that invisibility is blunt: if the same technology had been developed in the United States, everyone in the world would know about it. He frames it as a cultural gap rather than a technical one. Canada built something genuinely differentiated and then, with characteristic understatement, did not particularly insist that anyone notice.

The last new CANDU to enter commercial service was Cernavodă Unit 2 in Romania, in 2007. Almost twenty years have elapsed since then, and the Canadian domestic fleet has been in a sustained refurbishment cycle rather than a new-build cycle. That refurbishment program, covering units at Darlington and Bruce Power in Ontario, has been ahead of schedule and under budget. St. Julian argues that the refurbishments are the most important thing that has happened to the CANDU program in the last two decades because they generated the operational knowledge, supply chain capacity, and cost data that a new-build program requires.

The Metamorphosis of the Monark

The product AtkinsRéalis is now offering domestically is somewhat confusingly called the “Monark.” When AtkinsRéalis unveiled the Monark at the World Nuclear Exhibition in Paris in November 2023, the company broke with CANDU’s naming conventions entirely. Where previous designs had carried alphanumeric designations tied to their fuel channel count or development lineage, the Monark was positioned as something categorically new: a 1,000 megawatt single-unit design drawing on civil infrastructure of two designed but never built CANDU designs, the ACR-1000, the EC6, as well as the reactor core of the Darlington station, with CANDU’s passive safety features and a 70-year design life. Its marketing emphasized a break from the past.

That concept has since been substantially reoriented. What St. Julian described in our conversation is a configuration being referred to as the Monark-D, for Domestic. It consists of a strict four-unit replication of the Darlington station as it stands today, refurbished and relicensed, with as few material design changes as possible. The driver has come from the utilities who are becoming increasingly gun shy about first of a kind (FOAK) risk. That risk is increasingly being seen as the primary killer of nuclear cost and schedule. The way to bound it is to build what has already been licensed.

The confidence for replicating Darlington comes from its excellent operating experience and the confidence of the regulator. Darlington Unit 1 holds the world record for the longest continuous run of any thermal generating plant in history, 1,106 consecutive days, a figure made possible by CANDU’s online refuelling capability, which allows fuel replacement without shutdown.

The CNSC granted Darlington a 20-year operating licence in September 2025, the longest nuclear operating licence ever issued in Canada, following the completion of a C$12.8 billion refurbishment that finished four months ahead of schedule and C$150 million under budget.

The four-pack configuration matters economically: shared containment, a single control room, shared utilities, all of which pull the capital cost per megawatt below what any single-unit arrangement can achieve, which is partly why Ontario’s utilities have built this way since Pickering.

The Domestic CANDU Market

The near-term domestic pipeline St. Julian described is large. Bruce C, a four-pack on the Bruce Power site. Wesleyville A and potentially Wesleyville B, totalling another eight units east of Toronto. That is twelve reactors in Ontario alone, all potentially Monarks. Getting there requires first validating that the Darlington design, as built, can be crosswalked to current codes, standards, and regulatory expectations without requiring changes expensive enough to undermine the economics.

AtkinsRéalis is currently working through approximately 400,000 design documents: calculations, drawings, specifications, studies and re-stamping each one against today’s engineering standards with a professional engineering seal. The CNSC has identified 7 licensing areas out of a total of 18 that require examination given the evolution of codes since Darlington was originally licensed roughly forty years ago.

The three that St. Julian flagged as most technically involved are seismic criteria, aircraft impact requirements introduced after the September 2001 attacks, and the instrumentation and control (I&C) architecture, which mixes analog and digital systems in ways that must be demonstrated adequate against current safety standards.

On each of these, St. Julian’s position is that there are no fundamental licensing barriers, a term he was careful to distinguish from “no issues.” Fundamental barriers, he explained, are a concept that applies to genuinely novel reactor designs, where some core aspect of the physics or safety case has never been reviewed by a regulator.

CANDU has thirty-five operating units and a safety record the CNSC knows in detail. The question is not whether it is licensable. The question is whether satisfying any evolved requirements will require material changes expensive enough to shift the levelized cost of electricity (LCOE) beyond what Ontario’s utilities and ratepayers can absorb. AtkinsRéalis expects to have answers by the third quarter of 2026.

What will it cost?

While refusing to offer an on the record estimate, St. Julian’s core point is that repetition drives the marginal cost down, and CANDU has more repetitions behind it than any of its Western competitors in the current market. The AP1000 has two operating units at Vogtle in Georgia, and four in China with twelve currently under construction there.

The EPR has Taishan 1 and 2 in China, Olkiluoto 3 in Finland, Flamanville in France and Hinkley Point C still under construction in the United Kingdom.

There have been 35 CANDUs built but the raw global count is not the most telling number for the Monark-D specifically. The four-pack configuration that Darlington replicates has been built five consecutive times in Ontario alone: Pickering A, Pickering B, Bruce A, Bruce B, and then Darlington itself, twenty large multi-unit reactors in sequence, each generation absorbing the lessons of the last.

The Darlington units are effectively the 17th through 20th iterations of that configuration. The wrinkles accumulated across Pickering and Bruce, the steam generator designs, the I&C architecture, the containment geometry, the shared utilities layout, were progressively reworked before shovels broke ground at Darlington.

St. Julian was pointed about what this means for comparative LCOE calculations against competitors like Westinghouse now attempting to ramp the AP1000 from its 35 billion dollar cost for unit 1 & 2 at Vogtle to serial three: “I can’t reduce the cost from serial two to serial three by 50%. It can’t be done. It’s never been done in the history of the world.”

The refurbishment data reinforces this. The first Darlington unit to go through its mid-life refurbishment took 44 months from October 2016 to June 2020. Unit 4, completed in March 2026, was done in 968 days, roughly 32 months, four months ahead of its own schedule. From the first refurbishment to the last, the outage duration compressed by roughly 30 percent. The overall C$12.8 billion program landed C$150 million under budget.

The new-build Monarks are being designed with a 35-year initial licensed life rather than 30, and St. Julian projected that the first life extension following that period will cost roughly half what the Darlington refurbishments did, because the institutional knowledge will already exist in the supply chain and the workforce. A hundred-plus year asset with two economic life extensions, not a 35-year asset followed by a decommissioning decision.

The CANDU Export Market

Romania

The two CANDUs with the most direct bearing on international sales prospects are not in Canada. They are at Cernavodă in Romania. Four stand alone CANDU 6 units began construction there in the 1980s but were interrupted by the fall of the Ceausescu regime. Units 1 and 2 have been in operation since 1996 and 2007 respectively. However AtkinsRéalis is now under contract to finish the partially built Units 3 and 4 at the same site. The C6 design, 380 fuel channels, 730MW output, already licensed in Romania and certified by the European Union, is the product AtkinsRéalis is leading with in Eastern Europe.

Poland

Poland is the most advanced conversation. St. Julian and AtkinsRéalis CEO Ian Edwards were in Warsaw recently for a supplier day intended to introduce the CANDU supply chain to Polish industrial partners. Export Development Canada (EDC), the federal export credit agency, has received a request for $15 billion in financing support on the basis that Polish units would generate at least that much in Canadian content.

For context, the EDC committed $850 million in loan guarantees to the Cernavodă Unit 1 life extension and $3 billion to finish the partially built Units 3 and 4; AtkinsRéalis is currently revising that latter application upward to $4.5 billion as the front-end engineering design matures and Canadian content projections improve.

The pattern in these deals is consistent: EDC provides a low-interest sovereign loan to the buyer country, the contractual requirement for Canadian content means the money flows back into Canadian engineering firms and manufacturers, and the loan is repaid from electricity revenues over the life of the plant.

Turkey and Asean

Turkey is the other active conversation in Eastern Europe, somewhat further behind Poland. The Association of Southeast Asian Nations (ASEAN) countries represent a cluster of longer-horizon opportunities: St. Julian said AtkinsRéalis is in discussions with six of the eight or nine ASEAN members, none of which currently have nuclear power and all of which have begun developing domestic regulatory capacity.

China

China is the most complicated file. Canada has had a continuous presence there since the two CANDU 6 units were built at Qinshan Phase III, commissioned in 2002 and 2003 by AECL on a turnkey basis, on schedule and under budget by a team led by legendary project manager Ken Petrunik. Those units are entering refurbishment planning and early-phase execution through TQNPC (Third Qinshan Nuclear Power Company). Qinshan’s CANDU units are consistently among the top-performing reactors in China by capacity factor and reliability, with periods where they rank at or near the very top of the national fleet.

There were preliminary discussions about building two additional C6 units in part to boost China’s endogenous medical isotope production, over 80% of which occurs at the existing Qinshan units, before the geopolitical deterioration of Canadian-Chinese relations in the late 2010s closed them down.

The Advanced Fuel CANDU Reactor

The more technically interesting conversation concerns the Advanced Fuel CANDU Reactor (AFCR), a concept developed jointly by Candu Energy, CNNC, TQNPC, the Nuclear Power Institute of China, and China North Nuclear Fuel Company through a series of agreements from 2005 to 2016.

When a PWR discharges spent fuel, that fuel still contains roughly 0.9 percent uranium-235, compared to the 0.7 percent in natural uranium that CANDU was designed to burn. A PWR cannot reuse that material directly because even at the higher enrichment levels that went in, the discharged concentration is too low to sustain a PWR chain reaction. The uranium must first be recovered through conventional reprocessing, which chemically separates it from the fission products and plutonium accumulated during irradiation. That recovered uranium is then blended down with depleted uranium from enrichment tails to produce what the program calls Natural Uranium Equivalent fuel. Because CANDU’s high neutron economy can sustain a chain reaction at concentrations that would extinguish a PWR, this blended fuel feeds into a CANDU core without any re-enrichment.

The ratio the joint development work confirmed is 4:1: the spent fuel from four 1,000 megawatt PWRs can fully supply one 700 megawatt AFCR, extracting a further 30 to 40 percent of the energy remaining in that material, while simultaneously drawing down the stockpile of depleted uranium enrichment tails used as the diluent, a stream that is itself an expensive long-term storage liability accumulating at enrichment facilities worldwide. The detailed conceptual design was completed and the fuel technology was partially demonstrated at Qinshan. Twenty four NUE bundles were successfully irradiated in the demonstration program, however NUE fabrication remains an open engineering problem, with pellet sintering challenges that have so far required a more involved powder conversion process. Ultimately the political rupture in Canada-China relations ended the program.

To grasp why this matters to China specifically, consider the trajectory. China currently operates around 58 reactors totalling roughly 60 GWe, with 33 more under construction and credible projections putting the fleet at 300 GWe or beyond by 2050. China already imports 80 to 85 percent of its uranium requirements; at 300 GWe of predominantly PWR capacity, annual uranium demand would approach 50,000 tonnes in a global market that produced around 60,000 tonnes in total in 2023.

China’s domestic resources are limited relative to that scale, and the geopolitics of uranium supply, concentrated in Kazakhstan, Canada, and Australia, are not straightforwardly friendly to a country whose relationships with Western suppliers have been under sustained pressure. Theoretically at that fleet size, the spent fuel from China’s PWRs could theoretically keep 70 to 75 AFCRs running with no additional natural uranium input, taking a meaningful bite out of both the waste management problem and the uranium import dependency simultaneously.

The AFCR does require reprocessing infrastructure: the recovered uranium must first be chemically separated from fission products through PUREX or equivalent processing before it can be blended with depleted uranium tails. However, China is already building two reprocessing plants at 200 tonnes per year each, so that piece of the puzzle is in motion regardless.

The cost of NUE fuel is higher than fresh natural uranium and fabrication challenges remain, which is a genuine limitation, but China has consistently demonstrated that it is willing to pay a premium for reactor technologies that extend its strategic capabilities like the HTR-PM high-temperature gas-cooled reactor at Shidao Bay and CFR-600 sodium fast reactor at Xiapu.

The AFCR sits in a similar space: it does not minimize near-term electricity cost, but it partially closes China’s fuel cycle using a commercially proven reactor platform. Prime Minister Carney’s recent state visit to China included AtkinsRéalis in the delegation. The AFCR conversations, St. Julian said, are restarting.

Reviving the 1980’s Golden Age of CANDU

The domestic supply chain question is where the gap between ambition and execution is most concrete. Today approximately 100,000 workers are sustained by the Canadian nuclear industry, primarily by the Bruce and Darlington and now Pickering refurbishment programs. The refurbishments have been running for about fifteen years, which means the supply chain has had a longer warm-up period than most observers credit. St. Julian describes this as a significant advantage over competitors: “Canada is well ahead of, I call it, the power curve.” The Americans attempting to ramp up an AP1000 supply chain essentially from a standing start, and the French working to reconstitute capacity for EPR2, are trying to do something that Canada has been doing continuously, if at reduced scale, for a decade and a half.

The most important gap St. Julian identified in the CANDU supply chain is heavy water supply. Existing Canadian inventory can cover the first one or two new units by drawing on surplus from Pickering and Gentilly, but a twelve-reactor build-out in Ontario would require production capacity that does not currently exist in Canada. AtkinsRéalis has signed a memorandum of understanding with Argentina, whose state nuclear technology company designed heavy water plant, Planta Industrial de Agua Pesada (PIAP) at Arroyito has been in operation for roughly forty years and was taken offline about seven years ago; AtkinsRéalis is helping restart it, intends to take off-take from it, and sees the process design as licensable for a Canadian facility. Bruce Power, Ontario Power Generation (OPG), and AECL are all participating in the working group addressing this.

The recent thaw in Canada-India relations opens a further pathway. Prime Minister Carney’s March 2026 visit to New Delhi produced a landmark $2.6 billion agreement for Saskatoon-based Cameco to supply nearly 22 million pounds of uranium to India for its civilian nuclear program, alongside a new Strategic Energy Partnership and explicit commitments to cooperate on reactor technology.

India operates the world’s largest heavy water production complex and has decades of experience with CANDU-derived reactors. There is a satisfying narrative symmetry in the emerging relationship: Saskatchewan uranium fueling Indian reactors, Indian heavy water filling Canadian CANDUs.

Team CANDU?

St. Julian was direct about where the heavier political economy question sits. The US federal government has committed to support the AP1000 program through an $80 billion investment in Westinghouse and committing to assist with financing, streamline permitting, and facilitate construction. In exchange, the government receives a profit participation of 20 percent of any cash distributions above $17.5 billion, and holds the option to require a Westinghouse IPO by 2029 and convert that interest into a 20 percent equity stake. Separately, the DOE has designated four federal reservations: Oak Ridge, Savannah River, Paducah, and Idaho National Laboratory, sites originally acquired through eminent domain for the weapons program, as co-location zones pairing nuclear power generation with hyperscaler data centers as anchor load customers

The U.S. government is not simply financing the program; it is taking a position in the upside. Despite this federal commitment, Westinghouse has no active supply chain and limited construction capacity for the program it has just been handed.

Canada’s situation is structurally the opposite with an active supply chain and workforce. What Canada lacks is the political architecture that converts that industrial inheritance into a coordinated program.

St. Julian said AtkinsRéalis is in early discussions with both the federal government and provincial governments about whether something structurally analogous to the US model makes sense for CANDU, not an American-style program, he was careful to say, but some form of direct investment that would allow the Crown to benefit not just from GDP and tax receipts but from a greater return on the technology it originally funded.

Canada already participates in the upside from CANDU deployment, but it does so through an upstream mechanism tied to intellectual property rather than through direct project ownership or visible revenue sharing at the plant level.

When the federal government sold the CANDU reactor division of Atomic Energy of Canada Limited to SNC-Lavalin in 2011, it retained ownership of the core CANDU intellectual property. The commercial entity, now operating as Candu Energy under AtkinsRéalis, was granted a long term license to deploy and monetize that technology globally. In return, AECL receives royalties tied to new builds, refurbishments, and related engineering services. These payments are structured as royalties tied to future reactor sales, life extensions, and certain products and services, with the government projecting potential recovery in 2011 of up to $285 million over roughly 15 years though the precise payment formula has not been publicly disclosed.

The practical effect is that Canada captures value from every CANDU project, but indirectly. The government does not take an equity stake in individual plants, does not receive a per megawatt hour production royalty, and does not participate explicitly in project level cash flows. The return is embedded in licensing revenue, supplemented by domestic supply chain activity and tax receipts. The actual royalty receipts have never been publicly disclosed.

The contrast with the U.S. government Westinghouse deal is therefore structural. Canada operates as an intellectual property landlord, collecting a slice of revenue through licensing while leaving project economics to private developers and utilities. The emerging U.S. model places the state closer to the asset itself, aligning public returns more directly with reactor deployment and performance.

Getting the Band Back Together

In 1982 there were approximately 100,000 people working simultaneously in the CANDU supply chain, thirteen reactors under construction domestically and internationally at a single point in time, federal and provincial governments aligned with Crown utilities like Ontario Hydro, AECL as the technical authority, and a constellation of engineering firms and manufacturers operating in coordinated roles in a Team Canada approach.

That synergy has subsequently fractured over thirty-five years of privatization, restructuring, and new-build drought. AtkinsRéalis is now one of several large engineering firms, alongside Kinectrics, WSP, Stantec, and Hatch, where once AECL and Ontario Hydro were the organizing center. The construction contractors, Aecon, Bird, PCL, are capable but were not built around a continuous CANDU new-build program. The supply chain of 250 manufacturers exists, has been spooling up through the refurbishments, and in St. Julian’s phrase, is “fully capable” if not yet at the capacity a fleet build requires.

“We are in the early stages of getting the band back together,” he said. That being said the Team Canada framing is coherent only if Canada actually decides to field a team and play ball.

Picking that team, and backing it with the kind of structural commitment the US has made to AP1000, would mean converting a latent industrial advantage into an actual program. The raw materials for that program are more intact than the twenty-year new-build gap suggests. The refurbishments are the crawl: fifteen years of continuous work that kept the supply chain active, drove the cost curve down unit over unit, and gave the CNSC enough confidence in the refurbished Darlington design to license it for another twenty years. The Cernavodă contract is the walk: the first partial new CANDU construction since 2007, generating real cost and schedule data where before there were only projections. The domestic fleet build, twelve reactors in Ontario if the program goes the distance, is the run. Canada is approaching the transition from refurbishment to sustained new-build.

When new CANDUs break ground, the workforce stepping onto those sites will not be starting from scratch. The tools remain in place, and the institutional memory has been sustained through decades of refurbishment and ongoing engineering work. The supply chain now sees a credible pipeline of projects for the first time since the 1980s, which changes the nature of the challenge. Reassembling large scale build capacity becomes a problem of coordination and scaling, rather than reconstruction. Whether Canada chooses to mobilize these capabilities toward sustained deployment is the billion dollar question.

This essay accompanies Decouple’s conversation with Joe St. Julian, president of nuclear at AtkinsRéalis. Watch on YouTube, or listen on Apple and Spotify. If you find our work valuable, consider supporting Decouple by pledging on Substack or making a tax-deductible donation through our fiscal sponsor at Givebutter.

The energy crisis resulting from the War in Iran is bringing the catastrophic energy blindness of Taiwan’s political leadership into unforgiving focus.

Taiwan imports roughly 35% of its LNG from Qatar. LNG fuels nearly half of Taiwan’s grid after its political phaseout of nuclear power. The island maintains only about 11 days of LNG storage.

Had Taiwan kept its full nuclear fleet operating and commissioned Lungmen, its completed but never fuelled fourth nuclear plant, the country would today have roughly 7,750 MW of nuclear capacity producing about 61 TWh per year, covering around 21 percent of the grid.

Replacing that nuclear output with gas requires far more primary energy because Taiwan’s combined cycle gas turbines operate at roughly 55 percent thermal efficiency. Producing 61 TWh of electricity from gas therefore requires roughly 110 TWh of fuel input, equivalent to about 10 to 11 billion cubic metres of natural gas or roughly 7 to 8 million tonnes of LNG per year.

That volume is almost exactly the amount of LNG Taiwan currently imports from Qatar.

In other words, the nuclear fleet Taiwan shut down would have displaced essentially the entire Qatari supply stream.

Instead that capacity was retired and mothballed on political grounds and the gap was filled with gas.

In a fit of common sense, Taiwan held a referendum in August of 2025 on whether to restart the Ma’anshan nuclear plant, the island’s last operating reactor station, which had shut down in May after its forty year operating licence expired.

A clear majority of participating voters supported restarting the plant subject to regulatory approval and safety confirmation.

Taiwan’s referendum law, however, requires affirmative votes from at least one quarter of all eligible voters, roughly five million people. The referendum received about 4.3 million yes votes, leaving it below the legal threshold and keeping the plant offline, effectively confirming the continuation of Taiwan’s nuclear phaseout and LNG over dependence.

Taiwan depends almost entirely on above ground insulated LNG tanks at their import terminals. They function as a buffer designed for a world of uninterrupted deliveries rather than a strategic reserve designed to ride out a supply shock.

When a major node in the LNG system fails, there is no large fleet of idle ships ready to reroute, no spare liquefaction capacity waiting to fill the gap, and in Taiwan no underground storage that can stabilize supply while the market adjusts.

Taiwan’s nuclear shutdown therefore produced a structural vulnerability that is in the context of our current energy shock now impossible to ignore. The reactors that were closed would today be offsetting almost the entire volume of LNG Taiwan buys from Qatar.

There's never been a better time to restart Taiwan's nuclear fleet. Will Taiwan’s political class abandon their luxury beliefs and come to their senses?

Only time will tell.

When Japan shut down its reactor fleet after Fukushima in 2011 and scrambled to replace the lost generation with gas, Qatar delivered. When spot and short-term suppliers, including an Italian energy major and at least one trading house, voided their flexible contracts with Pakistan during the 2022 price spike and rerouted those cargoes to European buyers willing to pay more, Qatar delivered those too.

Then a $50,000 Iranian drone struck Ras Laffan.

The attack came in the broader wave of Iranian retaliation for the US-Israeli bombing campaign against Iranian nuclear and military infrastructure. Qatar, which had maintained some of the warmest relations with Iran of any Gulf state and had served as broker for the first Israel-Hamas ceasefire, was hit anyway.

The strike forced an evacuation of the facility and the first force majeure declaration in Ras Laffan’s history. The complex produces 77 million tonnes per year (Mt/y), roughly 19 percent of the 411 Mt traded globally in 2024. The North Field East expansion, the first of three planned phases that would nearly double Qatari capacity to 142 Mt/y by 2030, was approaching its first commissioning when the drone hit.

That same week, in the Mediterranean, a Ukranian drone boat struck the Arctic Metagaz, a Russian LNG carrier transiting toward the Suez Canal. The ship was not difficult to find. It was on a known route, visible on commercial vessel-tracking platforms, its name painted on the hull.

LNG tankers had already been rerouting around the Cape of Good Hope since the Houthi attacks made the Red Sea dangerous in late 2023, absorbing the extra transit time and cost rather than risk the strait. But no LNG carrier had been successfully struck. The Arctic Metagaz was the first. For Russia, losing one vessel from a fleet of sixteen shadow tankers is not a rounding error.

The two attacks came within days of each other, from different actors hitting different kinds of targets with different customers, but they broke the same assumption. The cascade that followed is not limited to European gas futures. It includes fertilizer plants in Pakistan going dark, rice farmers in Thailand who cannot get diesel, and nine days of reserve supply sitting in South Korean LNG storage tanks.

The Champagne of Fuels

To understand why the disruption hit so asymmetrically, it helps to understand what LNG actually is and why its supply chain concentrates risk the way it does. Natural gas, cooled to roughly minus 162 degrees Celsius, compresses to about one six-hundredth of its original volume, dense enough to ship economically across oceans.

That compression is the whole business: the liquefaction trains at Ras Laffan are the refrigerators, enormous industrial machines that chill gas into liquid and run continuously, taking weeks to months to restart after a cold shutdown; purpose-built cryogenic tankers are the thermoses, keeping cargo at minus 162 degrees Celsius across thousands of miles of open ocean; and the regasification terminals in Japan, South Korea, and Pakistan are the toasters that warm it back into pipeline-ready gas.

In a previous episode with Stephen we called LNG the champagne of fuels, and the analogy holds: it is expensive to produce, requires extraordinary handling at every stage, and arrives at the table only because an elaborate and costly infrastructure exists to get it there without incident.

Those thermoses are extraordinary machines in their own right. A single LNG tanker carries roughly one terawatt-hour of energy, the equivalent of around one month’s output from a gigawatt-scale nuclear plant, and each ship costs around $250 million to build. There are only about 700 in the global fleet, compared to over 7,000 oil tankers. But the fleet size is a symptom rather than the cause of LNG’s brittleness.

Every stage of the supply chain, from the liquefaction trains at the export terminal to the regasification plants at the receiving end, requires billions of dollars of fixed, specialized infrastructure that takes years to build and cannot be quickly replicated or redeployed when something goes wrong.

Oil markets have strategic petroleum reserves, idle tanker capacity, and a deep spot market precisely because decades of embargoes and supply crises taught them they needed those buffers. LNG has spent most of its relatively short history as a point-to-point business, the same ships on the same routes under decade-long contracts, operating in conditions stable enough that nobody was forced to build in the same shock absorption.

The storage picture compounds this vulnerability, and it splits sharply along geographic lines. Europe has geology on its side: depleted gas fields and salt caverns that can hold months of supply, which is why European utilities spend the summer methodically refilling underground storage to carry them through winter.

When Russia invaded Ukraine in February 2022, European gas prices spiked but did not immediately collapse supply, partly because storage cushioned the initial shock. The catastrophic price levels came later, in August and September, when utilities were competing to refill those same caverns for the following winter and the spot market could not keep up.

Asia has no equivalent. Japan, South Korea, and Taiwan depend on above-ground insulated LNG tanks at their import terminals, the same thermos logic applied onshore. South Korea had roughly nine days of LNG supply in reserve when Ras Laffan went down. Taiwan had around eleven days. Japan was similarly exposed.

These are not strategic reserves in any meaningful sense; they are operational buffers sized for a world of reliable, continuous point-to-point delivery, not for a prolonged outage at the source. When something breaks in the LNG chain, there is no large pool of idle ships to redeploy, no spare liquefaction capacity sitting ready, and in Asia, no caverns to draw down while the market finds its footing.

Ras Laffan in Qatar sits at the center of this system as its single largest node, supplying roughly one cargo in five to a global market that particularly in Asia has little tolerance for missed deliveries.

The Gas Field Qatar and Iran Share

Qatar’s rise is inseparable from the geology beneath it. The North Field, which Qatar taps for all of its LNG production, is the largest single natural gas reservoir in the world, sharing a geological boundary with Iran’s South Pars field across an invisible maritime line in the Persian Gulf. Because the reservoir is highly interconnected, aggressive extraction on one side draws down pressure on both. There is no formal bilateral agreement governing development.

Qatar’s 2005 moratorium was as much to appease Tehran as it was to safeguard the reservoir, and when Qatar lifted it in 2017, its energy minister announced that new development would target the southernmost portion of the field, furthest from the Iranian border. Iran’s lagging compression and reservoir management may have caused gas to migrate toward the Qatari side. Qatar has been, in effect, slowly draining a reservoir it nominally shares.

Through most of the 1970s and 1980s, nobody particularly wanted gas. Saudi Arabia simply burned its associated gas at the wellhead while running power plants on crude oil, at peak consuming as much as 900,000 barrels per day through its generators in summer.

Qatar sat on the world’s largest reservoir with no obvious market. The Japanese made that ecosystem possible: Japanese utilities provided the long-term purchase contracts that made financing of multi-billion dollar liquefaction plants bankable, and Japanese firms brought the cryogenic engineering that made the refrigeration systems work at scale. Qatar’s first export cargo departed in 1997, late to a game Brunei and Indonesia had been playing since the 1970s, but what followed lifted Qatar to the top of the LNG export rankings within a decade and transformed a country of fewer than 400,000 citizens into the wealthiest nation per capita in the region.

Then Fukushima accelerated everything: LNG imports rose by roughly 25 percent in the year following the accident, adding demand equivalent to an extra cargo arriving every day and a half. Qatar delivered, cementing the reputation for reliability that would make force majeure, when it finally came, so difficult to believe.

Iran watched from the South Pars side, unable to attract the technology partners, project financing, or long-term offtake agreements that sanctions had placed simultaneously out of reach. No major buyer would commit to two decades of purchases from a counterparty whose bank accounts could be frozen by Washington on 60 days’ notice. Iran cycled through Total, Linde, and CNPC, each of whom walked away when sanctions tightened or waivers failed to materialize.

The facility it began building in 2007 sat roughly 60 percent complete for over a decade, and Iranian gas went instead into a subsidized domestic economy where much was wasted and South Pars pressure maintenance consistently lagged, accelerating the migration of molecules toward the Qatari side.

The North Field East expansion, now under construction, was premised on a stable regional environment in which Iran would not be a hostile actor. The first phase was scheduled to come online by the end of this year. It is now under force majeure.

Why the Emerging World Bleeds First

The countries that feel the pain immediately are not the ones that dominate financial coverage of the gas crisis. South Korea, Japan, and Taiwan have enough money to go into spot markets and buy whatever is available, even at doubled prices. South Korea is subsidizing power prices to shield consumers; Japan is passing costs through to households. It is expensive, but survivable in the short term. For all three of those economies, the deeper vulnerability is their storage position.

Pakistan sources the overwhelming majority, approximately 99%, of its LNG from Qatar, a concentration rooted less in planning failure than in structural disadvantage. Pakistan entered LNG markets in 2015 with depleted domestic reserves and a sovereign credit rating so weak that no Western major would commit to long-term supply agreements.

Qatar was the only seller willing to take the country on which left Islamabad with almost no negotiating leverage. The 2016 agreements locked Pakistan into pricing at 13.37% of Brent, worse than the 12.66% India later negotiated, and included a take-or-pay clause and a Net Proceed Differential provision that captured any upside from resales for Qatar while leaving Pakistan exposed to losses.

The 2022 crisis then stripped away what limited diversification Pakistan had managed to assemble. Its contracts with an Italian major and a trading house evaporated as those cargoes were redirected to European buyers willing to pay crisis-era premiums. Qatar, constrained by its own long-term contractual obligations and its reputation as the market’s most reliable supplier, delivered. The lesson Pakistan drew was not irrational given the evidence in front of it: when the gas markets are strained, only Qatar shows up.

As Stapczynski explained, Pakistan’s biggest gas distributor has already cut supply to fertilizer plants. The same pattern is now running through Bangladesh and India, where roughly half of LNG imports have historically arrived from Qatar. Gas is the primary feedstock for urea synthesis, and when it stops flowing, the fertilizer supply chain seizes within weeks. Food price inflation follows within months. These are the second and third-order effects that an energy lens makes legible in ways that the images of strikes and counterstrikes do not.

The cascade of energy related impacts resulting from Qatar’s force majeure and the closing of the Straits of Hormuz is already visible across the region in sometimes seemingly mundane detail. China has ordered its largest refineries to halt diesel and petrol exports to protect domestic stocks.

India’s restaurants are warning of possible shutdowns as the government prioritizes gas supplies for households. Singapore’s bunker fuel providers, who supply the ships that carry everything else, are rationing supply. Thailand, where farmers in the north are already queuing for diesel with the rice harvest due in the coming weeks, is asking the broader public to limit air conditioning to 26 degrees Celsius, just as the country enters its hottest season.

That last detail matters more than it might appear. This conflict has arrived at the worst possible moment on the regional energy calendar. March through May is peak heat season across South and Southeast Asia, the period when gas-fired peaker plants carry the load for air conditioning demand, the way July and August do in the global north.

Replacement supply cannot be sourced from a spot market that has already been drained. The result, if this persists into April, will not be discomfort but blackouts, and they will fall hardest on the industrial zones and residential blocks where backup generation does not exist.

Share

The Shadow Fleet and the LNG Exception

One thread of the conversation that deserves particular attention is the vulnerability of Russia’s shadow LNG fleet, and what the attack on the Arctic Metagaz in the Mediterranean means for how energy warfare is evolving.

The oil dark fleet is large enough and diffuse enough to be difficult to fully suppress; there are simply too many ships, too many registry jurisdictions, and too many willing buyers for enforcement to be comprehensive. LNG is categorically different. Journalists like Stapczynski have counted approximately 16 shadow fleet LNG vessels operating for Russia, transporting sanctioned output from the Arctic LNG 2 facility that the Biden administration targeted in 2024.

When one of those ships burns, it is not a rounding error. Russia loses a significant fraction of its shadow export capacity overnight, and because LNG vessels are individually trackable by satellite image, the location of every remaining ship is known to anyone with a laptop and a subscription to a vessel-tracking service.

To understand why the shadow LNG fleet exists in its current form, it helps to understand the architecture of the sanctions regime it is navigating. The United States, European Union, and United Kingdom have all imposed measures targeting Russian LNG, most significantly the Biden administration’s 2024 sanctions on Arctic LNG 2 and, later, on Portovaya. But these are Western sanctions, and a substantial portion of the world does not recognize them as legitimate. China’s official position is that unilateral Western sanctions lack legal standing under international law, and Beijing has made no secret of its willingness to continue purchasing Russian energy.

How China Sidesteps Western Sanctions

The practical reality is more nuanced than a simple refusal to comply. Chinese companies with global operations, access to the US financial system, and exposure to dollar-denominated transactions have very strong reasons to avoid being designated under secondary sanctions, which can effectively cut a firm off from the entire Western financial architecture. The result is a bifurcation: the Chinese state openly rejects the sanctions framework while Chinese companies privately manage their exposure to it.

The Beihai terminal in Guangxi province illustrates the logic precisely. China designated Beihai as the sole entry point for cargoes from Arctic LNG 2, and other Chinese importers stopped using the terminal entirely. The terminal was chosen for its low-risk profile: it is located in China’s smallest gas-consuming region, meaning any potential supply disruption or sanctions enforcement would have limited implications for the wider market.

By concentrating sanctioned imports at a single, deliberately quarantined facility, Beijing effectively shielded the rest of its gas sector from secondary sanctions exposure. The ships supplying Beihai, all of them subject to US or EU sanctions, have largely abandoned any pretence of concealment, operating with their Automatic Identification System (AIS) transponders switched on. As one energy consultancy analyst put it, they realized they would be found out on satellite imagery within days regardless, so there was no point in hiding.

This is also why a single LNG tanker cannot simply rotate between Russian and American supply streams. Unlike a flag registry or a corporate owner, both of which can be changed through shell company transfers and reflagging maneuvers that are standard practice in the shadow fleet, a vessel’s International Maritime Organization (IMO) number is permanent.

The Office of Foreign Assets Control (OFAC) designation attaches to that number, which means a ship that has called at a sanctioned Russian facility carries that designation for as long as it remains on the Specialty Designated Nationals (SDN) list, and delisting requires active engagement with OFAC that almost never happens in practice.

Once designated, a vessel is effectively locked out of Western ports, Western insurance, and Western-adjacent supply chains, with its remaining market confined to buyers in jurisdictions that do not enforce US secondary sanctions. The major portfolio players, Cheniere, Shell, the Japanese utilities holding long-term US LNG contracts, cannot allow their supply chains to touch a designated vessel without triggering the secondary sanctions exposure they have spent considerable legal effort to avoid.

The market is not merely bifurcated by politics; it is bifurcated by the physical and legal incompatibility of the two supply chains, written permanently into a ship’s identity the moment OFAC adds its IMO number to the list.

The Attack on the Arctic Metagaz

The Metagaz attack, which Russia attributes to Ukraine, is the first confirmed successful strike on an LNG carrier. That fact will reconfigure insurance pricing, routing decisions, and the calculus of every operator who has been transiting these vessels through the Mediterranean toward the Suez Canal.

Since the Houthi attacks on Red Sea shipping began at the end of 2023, legitimate LNG operators had already been routing around the Cape of Good Hope rather than through the Bab-el-Mandeb strait. The shadow fleet had been using the Mediterranean-Suez route precisely because it was still open. That is no longer the case.

The ships supplying Beihai have since avoided the Mediterranean entirely, which forces them onto one of two routes: around the Cape of Good Hope, adding weeks to each voyage and compressing the economics of each delivery, or through the Northern Sea Route, which is where the constraints become genuinely acute.

Why the Northern Sea Route is not yet the Solution

The NSR is not simply closed in winter in the way that a road closes. It is navigable by degree, depending on ice class, icebreaker availability, and the specific conditions of a given season. The shadow fleet LNG tankers servicing Arctic LNG 2 are predominantly Arc4-class vessels, meaning they can handle light ice independently but require nuclear icebreaker escort for anything heavier. The Arc7-class icebreaking tankers, capable of navigating winter ice conditions that would stop an Arc4 vessel entirely, are in desperately short supply. for Arctic LNG 2.

Russia has been trying to close this gap domestically. Three completed Arc7 carriers sit undeliverable at a South Korean shipyard because Western sanctions prohibit their transfer, and Russia’s own Zvezda yard is building replacements, but slowly and with Chinese assistance substituting for the European technology now denied to it.

In the meantime Russia has floated the concept of a nuclear-powered LNG submarine carrier, a vessel that would be immune to surface ice conditions entirely. It sounds like science fiction, and it probably is for now, but it is the kind of idea that only gets seriously discussed when every conventional option has a serious constraint attached to it.

The end result is a fleet that is operationally viable in summer, strained in autumn, and severely degraded in winter, dependent on routes that are now either militarily dangerous or seasonally impassable.

What the Shale Revolution Actually Bought

The United States’ willingness to sanction Russian LNG facilities, to allow Israel to attack Iran, and to absorb the geopolitical risk of a Strait of Hormuz closure is only possible because of the shale revolution.

Even before the conflict with Iran the Biden administration’s decision to sanction Arctic LNG 2 and remove Russian supply from global markets rested explicitly on the US being large enough a producer to stabilize the market in the aftermath. That logic carried forward into Trump’s second term, which has been willing to absorb disruption that would have been politically intolerable when the US was an energy importer.

A United States that was still a major LNG importer, which is what the early 2000s looked like before hydraulic fracturing unlocked the Marcellus, Haynesville, and Permian gas plays, would have faced the same supply shock every other importing country is facing right now. As a result of the shale revolution the receiving terminals being built along the Gulf Coast to handle anticipated LNG imports were quickly converted into export terminals, though the word converted flatters the difficulty.

Running the process in reverse requires entirely different industrial equipment: the storage tanks, marine berths, and some pipeline connections could be reused, but the liquefaction trains had to be built from scratch on the existing footprint. In ten short years the US became the world’s largest LNG exporter, producing around 21 percent of global supply, neck-and-neck with Qatar before this crisis and positioned to capture significant market share during a prolonged outage.

That positioning insulates American households and industrial users from the acute price pain now hitting Europe and Asia. US LNG export contracts are indexed to Henry Hub rather than global spot benchmarks, which means the thirty-five to fifty percent surges recorded on European TTF do not transmit directly into domestic gas bills. The insulation is real however, it is not complete, and it narrows the longer the Strait stays closed.

Interestingly the pricing structure of US LNG contracts means that it is not the export terminal operators but rather the US’s trading counterparties, including China, who are positioned to profit most from the current tightness in the LNG market.

The Chinese Paradox

China’s position in all of this is the most structurally interesting, and it reflects decades of deliberate state planning around exactly this kind of scenario. On the oil side, China holds the largest onshore crude stockpiles in the world, estimated at 1.2 billion barrels as of January 2026, providing roughly 108 days of import cover at current refinery run rates.

Beyond reserves, China has spent years building supply routes specifically designed to bypass maritime chokepoints, deepening pipeline connections to Russia and Central Asia that are entirely immune to anything happening in the Strait of Hormuz. On the gas side, the insulation runs deeper still. China’s domestic natural gas production hit 246 billion cubic meters in 2024, sustaining roughly six percent annual growth for most of the past decade, with unconventional sources including shale, tight gas, and coalbed methane now accounting for 43 percent of that output.

China is the world’s largest LNG importer and is genuinely exposed to the supply shock, but less so than its regional neighbors because it holds assets they do not: the ability to pivot toward Russian pipeline gas and sanctioned Arctic LNG without the legal and reputational costs that would destroy a Japanese or Korean utility’s access to Western financial markets, and a coal fleet large enough to absorb demand that gas cannot serve.

There is also a contract dimension, where China’s position shades from paradox into something closer to structural advantage. Between 2021 and 2023 China signed over 20 million tonnes per year of new long-term contracts with US LNG projects, a volume Washington had hoped would bind Beijing into durable commercial dependency on American energy. Then in February 2025, as the tariff escalation accelerated, Beijing slapped a 15 percent retaliatory duty on US LNG imports and stopped taking physical delivery entirely.

Sinopec, CNOOC, Sinochem, and PetroChina began redirecting contracted US cargoes to Europe and other Asian markets, a move made legally straightforward by the destination flexibility written into most US LNG contracts. The other major portfolio holders, Japanese trading houses, Shell, Trafigura, Gunvor, and Vitol, were doing the same thing, but without the geopolitical irony.

By the time the Ras Laffan strike landed, China held large contractual claims on American gas production that it was actively routing away from its own shores and selling at a premium into the same European and Asian spot markets now scrambling for replacement supply.

Rather than being bound into energy dependency China has instead converted that dependency into a trading position, pocketing the arbitrage while keeping its own import portfolio oriented toward suppliers the United States has minimal ability to sanction.

The Refill Problem and the Summer Ahead

Gas markets have seasonality that does not always map onto the political timing of crises, which is why the timing of any resolution matters as much as the resolution itself. Stapczynski walked through the scenarios explicitly.

If the strait reopens and Qatar restarts tomorrow, the disruption is an expensive blip. These are not simple machines: bringing LNG trains back to operating temperature after a cold shutdown takes weeks to months even under ideal conditions, and Qatar has been sub-chartering out vessels it no longer needed during the outage, so reassembling the logistics of a dispersed fleet adds further delay. Qatar’s energy minister told the Financial Times as much in the first days of the crisis. The supply disruption will outlast any near-term diplomatic resolution, but the market absorbs it and moves on.

If the outage runs into June or July, the calculus changes entirely. The Northern Hemisphere refill season runs roughly from April through September, when European utilities pull large volumes from global spot markets to rebuild storage caverns for winter, precisely when Asian buyers are also increasing demand for peak cooling load. When Russia began reducing gas deliveries to Europe after its invasion of Ukraine in February 2022, prices spiked and then settled.

The catastrophic levels, Dutch TTF briefly above 300 euros per megawatt-hour, came in August and early September during the refill scramble. Morgan Stanley’s analysis, circulated in the first days of this crisis, suggested that an outage lasting more than a month would tip the LNG market into structural deficit for the year, meaning global demand would exceed available supply even with all other producers running at full capacity.

If the outage stretches into months, that scenario becomes the base case rather than a tail risk. The price discovery mechanism becomes brutal: the countries with the deepest pockets secure supply, and the countries without them go dark. Pakistan is still economically reeling from the 2022 energy crisis.

As Stapczynski put it, the way countries recover from energy shocks changes their trajectory permanently. Pakistan’s solar boom, its IMF bailout, its shift away from gas as an industrial input, all of it traces back to 2022. A repeat, at greater magnitude, with less warning, and arriving during peak heat season across South and Southeast Asia, is what the analysts are trying not to think about too carefully.

LNG after Ras Laffan and the Arctic Metagaz

The threshold for disrupting the global LNG market is no longer a functioning state with an air force or navy. Ras Laffan declared force majeure because of a thirty thousand dollar propeller driven drone. The Arctic Metagaz was destroyed by what was likely an unmanned surface drone small enough to be launched without conventional naval support.

LNG moves through an unusually concentrated system of liquefaction plants, storage tanks, and specialized carriers that cost billions of dollars and exist in only a handful of locations worldwide, so a small group with a grievance, a coastline, and access to technology that costs less than a used car can interrupt flows that underpin a large share of global gas trade.

The Israeli vision, insofar as it can be read from the targeting choices and the public statements of its strategic thinkers, is of an Iran so internally fragmented that it cannot project conventional military power outward. A state, perhaps resembling post-Assad Syria, broken along ethnic and regional lines, with Kurdish, Baloch, Azeri, and other peripheral groups empowered enough to prevent Tehran from reconsolidating control.

That vision may be coherent on its own terms. A fragmented Iran is less capable of supplying Hezbollah, less able to coordinate proxy networks across the region, and less able to place a ballistic missile on Tel Aviv. The global economy, however, now faces the prospect of a failed state alongside the world’s most critical energy chokepoint, and that may be a considerably harder problem to manage than a hostile pre war Iran ever was.

Before this war, Iran was a unified and largely legible actor. The Supreme Leader held final authority over the IRGC, the regular army, and the proxy networks simultaneously. That centralization was the thing that made diplomacy, however difficult, at least theoretically functional: there was a clear chain of command to negotiate with, a leadership whose equities could be threatened, and a foreign ministry capable of receiving and transmitting agreements. That engagement was attempted twice in recent memory through the nuclear deal framework and interrupted both times by US and Israeli surprise attacks before it could potentially produce durable arrangements.

What the bombing campaign has produced in its place is different. Iran has activated what it calls the Mosaic Defense doctrine, reorganizing the IRGC into 31 largely autonomous provincial units whose commanders no longer require approval from Tehran to launch missiles, drones, or maritime operations. Iranian Foreign Minister Araghchi acknowledged as much publicly, saying that certain units had become independent and were functioning according to previously issued operational guidance rather than continuous central direction.

The IRGC’s decentralization was designed as a military resilience strategy, developed after watching the collapse of centralized Iraqi forces in 2003. As an energy security problem it is something else entirely. Provincial coastal commanders with autonomous strike authority and staging points along the Persian Gulf present a categorically different threat to the Strait of Hormuz than a unified state does. A unified state can be deterred, negotiated with, and held to agreements. Thirty-one semi-autonomous military nodes distributed across Iranian territory present a more complex challenge.

A fragmented Iran that cannot threaten Tel Aviv may still be entirely capable of making the Strait of Hormuz ungovernable for commercial shipping, and unlike a state, it offers no foreign ministry to call, no supreme leader and no clear chain of command through which a ceasefire can be transmitted.

The affordable, reliable LNG that was supposed to serve as the bridge fuel for the developing world was always premised on freedom of navigation holding. The logic connecting these events is straightforward: the global energy system is built on infrastructure that is individually reachable, geographically concentrated, and defended by assumptions about acceptable behavior that have been visibly eroding for two years. Pakistan’s farmers, Thailand’s rice harvesters, and India’s industrial users are now on the front line of an energy war and caught paying the carrying cost of those assumptions now.

Share

Watch the full episode with Stephen Stapczynski on YouTube, or listen on Spotify and Apple Podcasts. If you find Decouple’s work valuable, consider pledging on Substack or making a tax-deductible donation through our fiscal sponsor at Givebutter.

For the “swiss watch” fuel of a technology often described as uniquely complex, that is a remarkable number. It reflects seven decades of industrial refinement, automated production lines running continuously to supply around 400 reactors worldwide, rejection rates under 1%, and a quality control regime so well understood that a single leaking fuel pin now constitutes a reportable anomaly rather than routine background noise.

Advanced reactor vendors selling their designs on narratives of lower construction cost, simpler containment, and inherent safety almost uniformly propose to achieve those outcomes through fuel that is more complex, more expensive, and manufactured by supply chains that do not yet exist at commercial scale. That observation alone does not invalidate their business cases, but it demands that the cost and supply chain assumptions behind their fuel be examined with the same rigour applied to their reactor construction schedules.

Uranium Dioxide: Good enough at everything without being very good at anything

Part one of this series based on my discussion with Michael Seely of the Atomic Blender channel covered how uranium dioxide (UO2) became the dominant commercial fuel. The short version is that it occupies an unusually comfortable middle position among candidate fuel materials: high melting point, reasonable fission product retention, chemical compatibility with zirconium cladding and pressurized water, and fabrication simple enough to automate at industrial scale. The process is direct: press the powder into pellets, fire them in a kiln, load them into rods, backfill with helium, and weld the ends shut. UO2 is not excellent at any individual property, but it is adequate at all of them simultaneously, and seven decades of production refinement have made that adequacy extraordinarily reliable.

Adequacy has limits. UO2 as a ceramic transfers heat poorly. The relatively poor thermal conductivity gap between the pellet’s outer surface and its center creates a temperature gradient of close to 1,000 degrees Celsius across a pellet roughly a centimeter in diameter. The centerline of a UO2 pellet operating in a pressurized water reactor (PWR) can reach 1,200 degrees Celsius while the coolant immediately outside the cladding sits at around 300. That margin sounds generous given UO2’s melting point near 2,800 degrees, but it shrinks rapidly as operating conditions change. In a high temperature gas reactor where coolant temperatures already approach 800 degrees, and where gas transfers heat to the cladding less effectively than pressurized water, the centerline temperature easily climbs toward 1,800 degrees. The 2,800 degree limit, which looked reassuring in a PWR, looks considerably less reassuring with 1,000 degrees already consumed by the operating conditions of the reactor itself.

UO2 also fractures under thermal stress. Fresh pellets loaded into a reactor are smooth and nearly geometrically perfect. After a few power transients, microscopically they resemble cracked clay: jagged sections separated by fracture planes that open release paths for fission gases into the gap between pellet and cladding. X-ray images of irradiated pellets show something unrecognizable as the original cylinder. The practical consequence is that the gap fills with fission gas under pressure driven by high temperatures, and that pressure constrains how aggressively the reactor can be maneuvered. This is manageable in a baseload PWR; it becomes a harder constraint in a reactor expected to follow grid demand through rapid power swings.

These are the problems that TRISO (tristructural isotropic) fuel and metallic fuel both attempt to address, through quite different approaches and with quite different implications for cost.

TRISO: A Brilliant Idea With an Expensive Manufacturing Problem

TRISO fuel is, at its core, still UO2. Understanding this matters because the technology is often described as a completely different paradigm, when it is more precisely a geometric rearrangement of the same fundamental material, surrounded by protective layers designed to solve the problems that scale and geometry create.

A TRISO particle consists of a UO2 kernel roughly the size of a poppy seed, surrounded by a porous carbon buffer layer, an inner pyrolytic carbon layer, a silicon carbide (SiC) layer, and an outer pyrolytic carbon layer. The carbon buffer exists to absorb the volumetric expansion and fission gas accumulation that would otherwise stress the surrounding layers; it functions essentially as a compressible void, a deliberate accommodation of the physics rather than a resistance to it. The SiC layer provides the actual structural containment. Silicon carbide at relevant temperatures has a melting point so high that even if the UO2 kernel inside were to melt, the SiC shell would remain intact. These particles are then embedded in a graphite matrix that provides moderation.

The thermal problem largely disappears at this scale. A poppy-seed-sized sphere of UO2 has a centerline-to-surface distance measured in fractions of a millimeter rather than half a centimeter; the temperature gradient across it is correspondingly small. The fission gas fracturing problem does not disappear entirely, but the particle’s geometry and the buffer layer substantially reduce its severity.

The performance case is real. TRISO fuel ran in the German AVR and THTR-300 reactors, in the American Peach Bottom and Fort St. Vrain plants, and their fuel behavior was generally good. Where those reactors encountered significant difficulties, the problems were usually in supporting systems rather than in the fuel itself. That historical record is the legitimate foundation of the enthusiasm around TRISO.

The manufacturing problem is where the gap between the technology’s physical elegance and its economic viability currently sits.

Producing TRISO starts with dissolving UO2 in acid to form a uranium nitrate solution, precipitating that solution into tiny droplets, and then drying and calcining those droplets into UO2 kernels. The kernels are supposed to be spheres. Many of them are not. The output resembles kicked balls, slightly dented, not perfectly round. Imperfect kernels are rejected, recovered, ground up, and run through the process again. The recovery of the high-assay low-enriched uranium (HALEU) feedstock from rejected material requires its own chemical deprocessing, because you cannot simply regrind an imperfect TRISO particle the way you can regrind a chipped UO2 pellet and put the powder back in the next batch.

The coating layers are applied by suspending the kernels in a gas flow and powder-coating them from the outside, layer by layer. Each layer must achieve uniform thickness across the entire surface of each kernel, across a production run of billions of kernels. Each layer has its own reject rate. Rejected coated kernels require additional chemical separation to recover the underlying HALEU before they can reenter the production stream.

The cumulative reject rates compound. A process that rejects 10% of kernels at the UO2 stage, another 10% at first coating, and another 10% at subsequent coatings does not produce a 10% overall rejection rate; the losses multiply through each step. The practical result is that current TRISO production, being conducted at laboratory and demonstration scale by BWXT, X-energy’s XTRISO facility, and Standard Nuclear (formerly associated with Ultra Safe Nuclear), involves small batches with considerable manual intervention, high reject rates, and expensive HALEU recovery operations at each failed step.

The cost consequence is predictable. HALEU feedstock at around 20% enrichment costs approximately $15,000 per kilogram, compared to about $3,000 per kilogram for conventional 5% enriched uranium. TRISO fabrication on top of that currently runs somewhere between $5,000 and $15,000 per kilogram, though these estimates carry wide uncertainty and the upper end reflects current small-batch production conditions rather than any plausible industrial future. Total fuel cost approaching $30,000 per kilogram is therefore not an unreasonable current and foreseeable future estimate, against the roughly $3,300 per kilogram for a conventional UO2 assembly.

However, it is reasonable to believe that cost will fall if TRISO demand grows. A transistor analogy is apt: microprocessors in the 1970s were expensive precisely because the manufacturing processes for very small, tightly toleranced structures had not yet been optimized through repetition at scale. The cost of computing fell by orders of magnitude as demand drove automation and process refinement. The analogy has a limit, though. Semiconductor demand exploded because chips unlocked entirely new categories of human activity that had no prior substitute. A TRISO-fueled reactor produces electricity and medium grade process heat, which the world already generates through dozens of competing technologies at lower cost and with more mature supply chains. The demand curve for a premium-fuel reactor selling into a commodity electricity market is not going to resemble the demand curve for the personal computer. The question for TRISO is therefore not just whether demand will materialize, but whether it will materialize in sufficient volume, and concentrated enough in a small number of fuel specifications, to drive the automation that would make the economics work.

At the moment, there is no operating TRISO-fueled reactor in the Western world. The only commercial TRISO reactors anywhere are the two units of the HTR-PM (High Temperature Reactor Pebble-bed Module) at Shidaowan in China, which achieved a capacity factor of around 20% in their second year of operation. Nine years of construction followed by a multi-year commissioning period, then a 20% capacity factor, in a country willing to absorb those costs as part of a broader national nuclear learning program. As a commercial proposition for private Western investors, that starting point requires considerable explanation.

Metallic Fuel: Better Physics, Harder Problems

Metallic fuel addresses different limitations than TRISO and has a different historical pedigree. Where TRISO is essentially a miniaturized containment vessel wrapped around familiar ceramic material, metallic fuel replaces the ceramic entirely with uranium in its metallic form, typically alloyed with zirconium.

The physical advantages are substantial. Metal conducts heat orders of magnitude better than ceramic. The centerline temperature problem that plagues UO2 effectively disappears; heat moves out of a metallic fuel pin quickly enough that the temperature gradient across the fuel is small, allowing operation at higher power densities without approaching melting limits. The metallic crystal structure also retains fission products more effectively than the ceramic grain structure of UO2, because displaced metal atoms tend to settle back into the crystal matrix rather than remaining in interstitial positions that open fission gas release paths.

The central disadvantage is that the melting point of uranium metal, and of uranium-zirconium alloys, is substantially lower than UO2. Less margin exists before the fuel melts if cooling is lost. The distinction between losing coolant flow and losing coolant entirely matters here. The Experimental Breeder Reactor II (EBR-2) at Argonne National Laboratory demonstrated in April 1986 that cutting the primary sodium pumps at full power, a loss of flow event, produced no fuel damage; the expanding fuel geometry triggered the reactor’s inherent negative reactivity feedback, reducing reactivity and dropping power to near zero within about five minutes, without any control rod insertion or operator action. The physics worked as designed, and it was a genuinely remarkable demonstration. What was not tested was a loss of coolant, a different and considerably more severe scenario for metallic fuel, where the lower melting temperature leaves less time between coolant loss and fuel damage than the large margin of UO2 ceramic would provide. The passive safety case for metallic fast reactors is real; it is also narrower than its most enthusiastic proponents sometimes suggest.

EBR-2 was also the site of a more ambitious vision, one that explains why the reactor exerts such a strong pull on the imagination of companies like Oklo today. The reactor itself sat beneath a silver dome at the Argonne National Laboratory site in Idaho, generating around 62 megawatts of thermal power. Physically attached to the side of this dome was the Fuel Cycle Facility, a reprocessing plant arranged in a U-shape around its processing equipment. The design intent was explicit: spent metallic fuel would leave the reactor, travel into the Fuel Cycle Facility, work its way through a sequence of electrochemical processing stations along the U, and return to the reactor as fresh fuel, all without leaving the site. The first irradiated fuel was reprocessed in the facility in September 1964, and recycled fuel returned to the reactor by April 1965. During the years the cycle operated, the turnaround time from fuel removal to reloading ran between 30 and 50 days.

The process itself operated nothing like the remote, automated image the word “reprocessing” might suggest. Spent fuel arriving from the reactor was intensely radioactive, requiring all handling to be done through thick leaded glass windows using remote manipulators, in rooms filled with inert argon gas rather than air, because both sodium bonding on the fuel and the highly reactive materials involved in the electrochemical steps would respond badly to oxygen or moisture. The electrorefining step dissolved the spent fuel into a molten salt bath of lithium chloride and potassium chloride at around 500 degrees Celsius, then applied an electrical current to selectively deposit uranium onto a steel cathode while fission products remained in the salt. The plutonium and higher actinides came out as a separate product mixed with rare earth fission products, deliberately left impure as a proliferation-resistance measure, since the mixture is unsuitable for weapons use without additional separation steps that would be detectable. The recovered uranium was then cast back into metallic fuel pins and returned to the reactor.

This is a compelling system in concept. It is also, in execution, one of the most demanding industrial processes ever attempted in a nuclear context.

Oklo, the company currently pursuing a commercial version of this concept, draws heavily on the EBR-2 legacy. Their proposed Aurora reactor, at 1.5MW in its original design and now discussed at larger scales, would use HALEU metallic fuel and eventually close its fuel cycle through on-site reprocessing of what they describe as legacy spent nuclear fuel from the existing light water fleet. It is a genuinely compelling narrative. It is also, for reasons outlined above, an ambitious one.

Oklo’s 2023 investor guidance described an initial core requiring approximately 10 metric tons of HALEU for a 50 megawatt reactor, lasting around 10 years before a reload of roughly half that quantity. The Centrus Energy HALEU demonstration cascade at Piketon, Ohio, currently the only Western HALEU enrichment operation, produces approximately 900 kilograms per year. A single Oklo reactor would therefore require roughly eleven years of Centrus’s entire output just for its initial fuel load. That ratio will improve if Centrus expands or other enrichment capacity comes online, but the scale of the gap between current Western HALEU production and even a modest fleet of advanced reactors is not a detail to be resolved by noting that it will eventually be addressed.

The fuel cost figure of $7,000 per kilogram cited in Oklo’s investor materials does not survive scrutiny. HALEU at market prices already costs roughly $15,000 per kilogram before a gram of it enters a fabrication facility. Achieving $7,000 per kilogram as an all-in fuel cost would require either a negotiated subsidy from the Department of Energy, uranium prices significantly lower than current market rates, or an assumption about reprocessing economics that the existing evidence does not support.

Reprocessing is not a simple recovery operation. French LWR reprocessing at La Hague has consistently failed to compete economically with fresh uranium at current market prices. The gap is wide enough that EDF, the state utility operating France’s nuclear fleet, has carried its stockpiles of recovered plutonium and reprocessed uranium at zero book value since 1995, a formal accounting admission that these materials cannot be sold at any price that covers the cost of turning them into usable fuel. France sustains La Hague through deliberate energy policy, not because the economics compel it.

CERMET: Combining the Best of Both Worlds

The ceramic metallic (CERMET) fuel concept exists because both UO2 pellets and pure metallic fuel involve real tradeoffs that neither alone fully resolves. The basic architecture embeds ceramic fuel particles, typically uranium dioxide, in a metal matrix rather than pressing them into solid pellets and stacking them in a rod. The metal matrix conducts heat far better than a solid ceramic pellet, which reduces the centerline temperature problem; the ceramic particles retain fission products better than a homogeneous metallic pin, because the particle boundaries impede fission gas migration. Neither property is fully optimized, but both are meaningfully improved over either pure form, and the combination opens a design space that conventional pellet fuel cannot access.

Russia has been operating CERMET fuel in its nuclear icebreaker fleet continuously since the 1950s, accumulating more than 400 reactor-years of operational experience across successive reactor generations. The current RITM-200, developed by Afrikantov OKBM and now installed on four icebreakers with more under construction, uses a cassette-type core of 199 fuel assemblies containing UO2 particles dispersed in a silumin matrix, an aluminum-silicon alloy, enriched to just under 20% uranium-235. The silumin matrix conducts heat substantially better than the zirconium-based structures of conventional light water reactor fuel, reduces centerline temperatures, and tolerates the swelling that accumulates during a long fuel residence without the cracking behavior that plagues UO2 pellets under power cycling. The fuel rods are clad in a nickel-chromium corrosion-resistant alloy rather than zirconium, chosen specifically for its performance over the long refueling cycles the icebreaker mission demands. The RITM-200 core operates for 5-7 years between refuelings, in reactors that must handle continuous, aggressive power transients as the icebreakers throttle up against pack ice and back off in open water. That duty cycle would be punishing for conventional UO2 fuel. The CERMET matrix handles it routinely.

The operational significance of this is worth considering. The RITM-200 is not a laboratory demonstration or a national prestige project running at artificially low capacity to generate data. It is working propulsion equipment on vessels breaking ice in the Arctic, operated by crews who cannot easily return to port if a fuel problem develops. The icebreaker experience proved the fuel concept under conditions more demanding than any grid-connected power reactor will ever face.

The reasons this experience receives relatively little attention in Western advanced nuclear discussions are not entirely mysterious. Russian reactor data is difficult to access and published under terms that make independent verification hard. The RITM-200 fuel is manufactured by Russian industrial infrastructure with no Western equivalent, making it a poor reference point for companies trying to establish domestic supply chains. And the enrichment level, just under 20%, sits at the boundary of what Western nonproliferation policy treats as acceptable for civilian use, which adds a layer of political complexity to direct comparisons. None of these factors diminish the technical significance of what Russia has demonstrated, but they explain why it tends to appear as a footnote in discussions dominated by American and European vendor narratives.

Lightbridge Corporation is pursuing a different application of the metallic fuel concept, one aimed at the existing light water reactor fleet rather than at next-generation designs. Their cruciform-shaped metallic fuel rod, developed over more than a decade and now at the test rod stage at Idaho National Laboratory, is designed to replace conventional UO2 assemblies in operating pressurized water reactors without requiring HALEU. The helical cruciform geometry increases the surface area available for heat transfer and improves coolant mixing around the rod, which are the same principles that underlie the Russian icebreaker fuel geometry. At conventional enrichment levels, Lightbridge argues that the metallic matrix’s better thermal conductivity and fission product retention could enable higher power density, higher burnup, and longer operating cycles than current UO2 assemblies allow, improving economics for existing reactors without requiring new reactor construction.

The proposition is commercially clever because it targets an installed base of around 400 operating reactors rather than a fleet that does not yet exist. The question that irradiation testing at commercial scale will eventually have to answer is whether the performance improvements are large enough to justify the cost and regulatory burden of qualifying a new fuel form for reactors licensed around the behavior of pellet fuel, and whether those improvements can be maintained against the continuing refinement of conventional UO2 fuel through accident-tolerant fuel programs and LEU Plus enrichment developments. A fuel that is better than today’s UO2 may still lose the economic competition if the target is moving.

The Variable That Actually Drives Outcomes

Advanced reactor companies routinely project capacity factors in the low to mid 90s. This is however the performance level of a mature, well-operated light water reactor fleet after five decades of systematic improvement, not the performance level of light water reactors in their first decades of operation.

The 93% capacity factor that currently anchors American light water reactor economics accumulated through a specific institutional process: hundreds of reactors generating comparable operating data, an industry-wide information sharing structure through the Electric Power Research Institute (EPRI) and the Institute of Nuclear Power Operations (INPO), sustained regulatory attention to failure modes, and multiple fuel vendors competing on reliability as much as price. That process took decades and required a fleet large enough to generate statistically meaningful data across failure modes that individual plants encounter only rarely.

An advanced reactor running on TRISO or metallic fuel will not arrive at 93% capacity on its first cycle, just as the Shidaowan HTR-PM did not arrive there in its second year. It will encounter failure modes that simulations and laboratory testing cannot fully predict, because the interaction between fuel behavior, reactor conditions, and supporting systems only becomes fully legible under real operating conditions. The cost of learning those lessons is real, and it falls on whoever finances the early units.

The fuel cost question and the operational performance question are not independent. A reactor running at 20% capacity factor with expensive HALEU fuel is extracting a small fraction of the potential revenue from a very expensive input material. The economics of advanced nuclear do not become compelling simply because the reactor construction cost comes down, if the fuel cost goes up and the capacity factor starts low and climbs slowly through costly operating experience.

None of this means TRISO or metallic fuel are dead ends. The historical performance of TRISO in its operating demonstrations was generally good. Metallic fuel’s physical properties are genuinely superior to UO2 in several respects that matter. CERMET fuel’s performance in Russian icebreakers is operationally proven in a demanding environment. The question is the sequencing: how long does the learning process take, how much does it cost, and who pays for it?

China can absorb the cost of the Shidaowan experience because it represents a small fraction of a state-directed nuclear program whose ambitions extend far beyond any single technology. A Western venture-backed company, or even a consortium of industrial partners with hyperscaler power purchase agreements behind them, will face a more compressed timeline between first operation and the need to demonstrate commercial competitiveness. The fuel economics are a central part of that demonstration, and the current numbers, on both cost and supply chain availability, are demanding.

This is the structural question that runs underneath all of it: what allowed water cooled reactors to become self-sustaining as a commercial proposition, and why has no other reactor type crossed the same line? The answer is not that water cooled reactors were superior on every technical dimension. They were not. Sodium-cooled fast reactors breed fuel. Gas-cooled reactors achieve better thermodynamic efficiency at high temperatures. The relevant distinction is not which technology had the most attractive properties on paper, but which technology cleared the hurdle of organic private investment, and why.

Light water reactors cleared that hurdle because the hurdle itself was not that high. The technology built on two centuries of industrial experience managing pressure vessels and steam as a working fluid. Enrichment services, zirconium cladding supply, UO2 pellet fabrication, vendor warranties, insurance pricing, regulatory precedent: all of this infrastructure consolidated around light water requirements during the 1960s and 1970s, under conditions where government bore much of the early risk. By the time utilities were making investment decisions on their own balance sheets, they were buying into an existing system rather than funding the creation of one. The industry could sustain the technology organically, without a state apparatus continuously absorbing the burn-in costs of first-of-kind systems running well below their projected capacity factors.

No alternative reactor technology has yet crossed that line. Non-LWR concepts still require some combination of a novel fuel supply chain, reprocessing infrastructure, specialized materials expertise, or regulatory precedent that does not exist at commercial scale. An advanced reactor vendor is not simply selling a reactor; it is implicitly asking early customers to co-finance the creation of the industrial ecosystem the reactor depends on. X-energy cannot deploy its Xe-100 without first building out the TRISO-X fabrication facility in Oak Ridge, a dedicated HALEU TRISO production line that has no customers until the reactor fleet exists and no economic rationale without the customers. Oklo cannot close its fuel cycle without a commercial-scale pyrochemical reprocessing capability that the Western world has never operated outside a national laboratory. Kairos Power’s fluoride salt-cooled reactor depends on TRISO pebbles produced to its own geometry and specification. Terrestrial Energy’s molten salt design requires a fuel form that no commercial fabricator currently produces.

Each of these supply chains has to be invented, licensed, capitalized, and staffed before the reactor it serves can operate at anything approaching commercial scale, and each is essentially useless to any other concept in the current landscape. That last point compounds the problem considerably. Where the light water fuel cycle achieved its economics partly through consolidation around a small number of converging fuel specifications, the current generation of advanced reactor developers has produced a proliferation of bespoke fuel types so different from one another that shared infrastructure is nearly impossible to justify. The result is that the fixed costs of building each supply chain get amortized over whatever modest fleet a single vendor can assemble, rather than across an industry, which is precisely the condition that keeps unit costs high and prevents the kind of learning-by-doing that made LEU Uranium dioxide fuel fabrication affordable.

The conventional fuel cycle became boring through decades of unglamorous industrial learning. Advanced fuel cycles will have to earn their own version of boring.

That process is starting now, at small scale, in demonstration facilities and national laboratory test reactors, and the timeline for it to reach industrial maturity is measured in decades rather than years. Acknowledging that clearly, rather than projecting mature performance onto first-of-kind systems, is the honest foundation for evaluating where advanced nuclear actually goes from here.

This essay accompanies part two of Decouple’s deep dive on the nuclear fuel cycle with Michael Seely. Part one, on conventional UO2 fuel, fabrication, burnup limits, and accident-tolerant fuel, is available now on the Decouple Substack. Watch on YouTube, or listen on Apple and Spotify.

If you find our work valuable, consider supporting Decouple by pledging on Substack or making a tax-deductible donation through our fiscal sponsor at Givebutter.

Nuclear fuel does almost none of that. The fuel elements that go into a reactor and the fuel elements that come out are, to a first approximation, the same material in the same geometry, sitting in the same place. A spent fuel assembly pulled from a reactor after six years of operation looks nearly identical to the fresh one that went in. The mass has changed by a tiny fraction of a percent, nuclear alchemy has occurred in which half the periodic table has been generated in the form of fission products within the ceramic pellets but the volume and geometry is essentially identical.

This one fact, that nuclear fuel must be preserved rather than destroyed, that the job of every layer of every system surrounding the core is to maintain the integrity of a material through years of radiation bombardment and extreme temperature gradients, shapes much of nuclear engineering. It explains the cladding materials, the obsessive quality control in fabrication facilities, and the decades of slow, painstaking improvement that have transformed a fleet that routinely operated with failed fuel elements into one where a single leaker triggers a formal investigation.

I spent a long conversation with Michael Seely, a nuclear enginner with a focus on fuel and the mind behind the Atomic Blender channel, going through what nuclear fuel actually is, how it is made, why it fails, and how the industry learned to prevent those failures. What follows is my attempt to synthesize that conversation into something useful for anyone who wants to understand nuclear from the inside out.

What the Fuel Actually Is

The commercial nuclear fuel cycle, in its conventional form, converges on a single material: uranium dioxide, or UO2. Regardless of reactor type, whether you are talking about a pressurized water reactor in France, a boiling water reactor in Japan, or a CANDU in Ontario, the fuel pellet sitting at the centre of the fuel rod is almost certainly a dense ceramic cylinder of uranium dioxide roughly the size of a fingertip.

UO2 ended up in this position for reasons that are easier to appreciate once you understand what you are asking a fuel material to do. You need something that can withstand centerline temperatures of 1,200-1,600 degrees Celsius under normal operating conditions, while the coolant immediately outside the cladding sits at around 300 degrees, a gradient of nearly a thousand degrees across a pellet roughly a centimetre in diameter.

You need something that will not chemically react with zirconium cladding or the pressurized water flowing over it. You need something that will trap the fission products, the gases and solids generated as uranium atoms split, inside its crystalline matrix rather than releasing them into the coolant. And you need something that can be manufactured reliably, in quantity, at a cost that keeps nuclear electricity commercially competitive. In fact the key differentiator between nuclear and fossil power generation is that despite its complexity nuclear fuel remains a relatively very small contributor to operating expenses.

Uranium dioxide satisfies all of these requirements tolerably well, which is distinct from satisfying any of them perfectly. It is a ceramic, which means it has an extremely high melting point, around 2,800 degrees Celsius, providing enormous safety margin even under severe accident conditions. Its crystalline grain structure traps fission products reasonably effectively: the krypton, xenon, and iodine gasses generated by fission mostly stay embedded in the UO2 matrix rather than migrating into the gap between pellet and cladding. And the manufacturing process, while technically demanding, has been refined over seven decades into something industrial routine.

The fabrication sequence is instructive precisely because of how unglamorous it is. Enriched uranium hexafluoride arrives at the fabrication facility as a gas. It is converted to UO2 powder through either a wet or dry chemical deconversion process, both involving steam or water stripping the fluorine atoms away and producing hydrofluoric acid as a byproduct, which is why fuel fabrication facilities will let you tour almost everything except the conversion and deconversion sections. The resulting powder is pressed into pellet shapes roughly double the final size, then sintered at high temperature in a controlled atmosphere, where the individual powder grains fuse together and the pellet densifies and shrinks to its final dimensions. It gets inspected, polished, and loaded into zirconium-alloy tubes that are backfilled with helium and welded shut. Those rods are then assembled into fuel assemblies and shipped to a reactor.

That is how you make nuclear fuel. The process has no truly exotic steps. Given adequate equipment and trained staff, it is not categorically harder than pharmaceutical tablet manufacturing or precision ceramics production for industrial applications. The difficulty is not in any single step but in the aggregate quality requirements, and in the fact that the finished product has to perform flawlessly inside a reactor for six years before it is seen again.

The Bespoke Problem

One of the things I had naively assumed before talking to Michael was that light water reactor fuel had converged on something close to a standard. If you are making a Westinghouse pressurized water reactor, you use Westinghouse fuel, and presumably that fuel works in any other Westinghouse plant. The reality is considerably messier.

At a high level, yes, the fuel concept is consistent: enriched UO2 in a zirconium-clad rod assembled into a square grid. But the geometry, enrichment, dimensions, grid structure, and pin layout vary substantially across reactor families and, within those families, often across individual plants depending on their vintage, their operating history, and the specific optimization choices made by the utility and fuel vendor over decades of operation. Westinghouse plants run different fuel from Framatome plants. BWRs, which have gone through generations of lattice designs from 6x6 grids to 10x10, use different fuel from PWRs. Even within a single family, the axial enrichment profile, the radial enrichment distribution across the assembly, and the height of the active fuel zone will vary between plants.

The reason for this complexity is worth understanding, because it is not bureaucratic perversity. Core designers are trying to solve an optimization problem of extraordinary dimensionality. A typical PWR core has 150 to 200 fuel assemblies, each containing more than 200 individual fuel pins, each pin containing hundreds of pellets, and each of those pellets will spend up to six years inside the reactor core before being removed. Over that period, neutron flux, fission product buildup, cladding corrosion, and mechanical stress all interact in ways that depend on where in the core the assembly sits, how many cycles it has already completed, and what the surrounding assemblies are doing. The enrichment of each pellet is therefore not a single number decided once and applied uniformly; it is a three-dimensional distribution optimized to extract maximum energy while staying within safety limits and protecting the structural components of the reactor vessel from excess neutron bombardment at its ends.

Michael’s description of axial enrichment zones, where pellets at the top and bottom of a fuel rod are enriched to lower levels to reduce neutron fluence on the reactor pressure vessel, was one of those small details that makes the whole enterprise feel astonishingly sophisticated once you register it. The pellets in the center of the rod might be enriched to 4.5 or 5 percent uranium-235. The pellets near the ends might be at 2 percent, or even natural uranium, deliberately and precisely placing less fissile material where you want less activity. In a BWR, where steam voiding changes the neutron moderation environment as you move up the core, the axial enrichment zoning becomes even more granular, potentially spanning seven or eight distinct levels.

The practical consequence of all this customization is that fuel is not a commodity. You cannot easily substitute one vendor’s assembly for another’s in an existing core, and a fuel design qualified for one plant is not automatically qualified for another even if they share a vendor and a general reactor family. Qualification is expensive and slow, which is one reason why GE Hitachi’s sales pitch for the BWRX-300 being constructed at Darlington explicitly highlights the use of existing GNF2 boiling water reactor fuel. Bringing in qualified, proven, extensively documented fuel avoids years of regulatory work and removes one of the largest uncertainties in the licensing process. It is also, as Michael pointed out, one of the more underappreciated arguments for fleet standardization: if you are building ten AP1000s rather than ten bespoke reactor designs, you eventually reduce to something approaching a standard fuel specification, and the fabrication, qualification, and supply chain costs all follow accordingly.

Leakers, Zero by 2010, and the Patience of Industrial Learning

The history of fuel failures in the American light water fleet is a useful corrective to the assumption that nuclear technology arrived mature and simply had to be deployed. Through the 1970s, 80s, and well into the 90s, it was routine for a majority of the US fleet to operate with fuel pins that had developed small holes, what the industry calls leakers. Fission gases would escape through these holes into the reactor coolant system, radiation monitors would detect the activity, and operators would respond with compensatory measures ranging from reducing power in specific assemblies to adjusting coolant chemistry. The reactor kept running. The leakers were noted, managed, and tolerated as a fact of life.

The causes were mostly prosaic. Debris ingress was the dominant mechanism: metal filings, wire bristles from cleaning brushes, small hardware left behind during maintenance, any of which could get lodged between fuel rods and vibrate its way through the soft zirconium cladding over months of operation. Weld defects at the end plugs of fuel pins could allow water ingress, triggering a hydrating reaction that embrittled the cladding. Grid-to-rod fretting, where the structural grids holding the assembly together wore against the fuel pins under flow-induced vibration, produced a class of failures that took years of systematic investigation to understand and address.

The economic consequences were real. In boiling water reactors, where steam goes directly to the turbine rather than through an intermediate steam generator, fission product contamination of the turbine island created radiation exposures that complicated every subsequent maintenance outage, requiring additional shielding, extra radiation surveys, and prolonged cleanup. In both BWR and PWR plants, a fuel failure meant elevated coolant activity for the remainder of the cycle, which meant higher worker dose during the outage, stricter controls on component removal and handling, and sometimes early cycle termination if activity climbed high enough. These costs accumulated in outage labor hours and replacement power purchases, which is why the industry eventually decided to treat them as urgent problems worth solving rather than conditions to be managed.

The Zero by 2010 initiative, organized through the Electric Power Research Institute and involving coordinated action across the US fleet and its fuel vendors, targeted exactly this. Foreign material exclusion programs were tightened at every stage of the fuel fabrication and reactor maintenance process. Grid and fuel rod designs were modified to reduce fretting susceptibility. Welding and inspection procedures were revised. And, critically, operators started sharing failure data systematically across the industry rather than treating each incident as a local problem to be quietly resolved.

By 2010, the initiative had largely succeeded. The US fleet today operates with fuel failure rates that would have been unimaginable in 1985. When a leak does occur now, it is investigated as an anomaly, not absorbed as background noise. The Taishan EPR incident that drew international attention a few years ago, where elevated primary circuit activity pointed to a small number of failed fuel rods, was notable precisely because fuel failures are no longer routine; the detection thresholds are sensitive enough and the expectations high enough that a handful of leakers in a fleet of hundreds of reactors generates headlines.

This history is worth holding in mind every time someone talks about the 93 percent capacity factors that today’s light water fleet achieves and then projects that number directly onto whatever new reactor concept and fuel they are discussing. The 93 percent is the product of decades of systematic learning, investment in understanding failure modes, and industry-wide coordination of the kind that only becomes possible once you have hundreds of units in operation generating comparable data. It was not the starting point. It was the destination of a very long journey.

Burnup, Cladding, and Why Things Changed After Fukushima

Burnup is the nuclear engineering metric for how hard you have used your fuel: how much energy you extracted per metric ton of uranium loaded. CANDU reactors running on natural uranium, where the tiny fraction of fissile uranium-235 is exhausted relatively quickly and the buildup of neutron-absorbing fission products forces fuel changes after a short irradiation period, achieve around 7 to 8 gigawatt-days per tonne. Light water reactors, starting with enriched fuel and benefiting from the in-situ generation of fissile plutonium from the abundant uranium-238, run to 50 or 60 gigawatt-days per tonne today, with higher burnup fuels pushing toward 70 or 80.

The limiting factor at high burnup is not purely enrichment level or the fuel pellet itself. The UO2 can in principle sustain continued fissioning as long as fissile material remains and the neutron environment supports it. What gives out is the zirconium cladding surrounding the pellet. Two degradation mechanisms are the main concerns. Oxidation of the cladding’s outer surface by the hot pressurized water, carrying dissolved boric acid in PWRs, progressively converts zirconium metal to zirconium oxide and depletes the cladding wall thickness. Hydrogen pickup, where hydrogen liberated by the oxidation reaction diffuses into the cladding metal, forms zirconium hydride precipitates that reduce the material’s ductility, making it brittle and prone to cracking under mechanical stress. Under normal operating conditions these processes are gradual and manageable. Under accident conditions, particularly rapid temperature transients, the embrittled cladding can fracture catastrophically.

Fukushima refocused the industry’s attention on this second concern. The hydrogen generated by rapid oxidation of zirconium cladding under loss-of-coolant conditions was directly responsible for the building explosions at Fukushima Daiichi. The response was accident-tolerant fuel, or ATF, an umbrella term for fuel designs intended to reduce the rate of cladding oxidation and hydrogen generation under severe accident conditions, buying more time for operators and emergency systems to respond before the situation becomes irretrievable.

The most commercially advanced ATF approach involves chromium-coated cladding, where a thin chromium layer on the outside of the zirconium tube dramatically slows oxidation at high temperatures. Doped fuel pellets, where small additions of chromia or alumina to the UO2 matrix improve grain structure and reduce fission gas release, complement the improved cladding. The combination, being developed commercially by both Framatome and Westinghouse, does not eliminate hydrogen generation under the most severe scenarios, but extends the coping time meaningfully.

What emerged from the development work, somewhat unexpectedly, was that the same improvements enabling better accident performance also enable higher burnup under normal operating conditions. A more oxidation-resistant cladding that retains ductility at higher burnup, combined with a fuel pellet that holds its geometry and retains fission gases more effectively, opens the door to fuel enriched to 6, 7, or 8 percent uranium-235, well above the conventional 5 percent ceiling. This is LEU Plus, and its commercial appeal is not primarily about accident safety, it is about outage economics.

A reactor that can run 24 months between refueling outages instead of 18 eliminates one scheduled outage every three cycles. At a plant like Vogtle, where four units now share a site and maintenance staff, being able to stagger outages cleanly on a 24-month cycle instead of overlapping 18-month cycles reduces the requirement for contractor surge staffing and simplifies the planning horizon considerably. Fuel that costs two to three times more per kilogram but that you reload less often and from which you extract more energy can be economically rational, especially given that fuel represents only 15 to 20 percent of a nuclear plant’s operating cost. Whether LEU Plus becomes standard across the fleet or remains a premium option for specific operating configurations will depend on plant-by-plant economics, but the direction of travel will be intriguing to watch.

The Chicken and the Egg of the HALEU Supply Chain

All of this background on conventional fuel provides essential context for understanding why high-assay low-enriched uranium (HALEU), the feedstock required by most advanced reactor designs, represents such a significant challenge.

HALEU is enriched to between 5 and 20 percent uranium-235. That range might sound modest relative to the 90 percent enrichment of weapons-grade material, but the difference in manufacturing infrastructure between 5 percent and 19.75 percent is enormous. At 5 percent, the entire global enrichment and fuel fabrication industry has been optimized for decades, with equipment geometries, material accounting procedures, criticality safety protocols, and facility designs all calibrated to that enrichment level. Moving to 19.75 percent changes essentially all of these parameters. Batches must be smaller to stay subcritical. Water must be kept further away. Accounting must be tighter. The facilities look different, operate differently, and require different trained staff.

The economics compound the infrastructure challenge. Making one kilogram of 5 percent enriched uranium requires about 9 kilograms of natural uranium as feedstock depending on tails, plus a well-understood amount of centrifuge separative work. Making one kilogram of HALEU at 20 percent requires roughly 40 kilograms of natural uranium and about six times the separative work per kilogram of product. HALEU is therefore expensive before a gram of it enters a fabrication facility, and the fabrication of HALEU into finished fuel is a separate challenge with its own cost premium.

The demand side compounds things further. The entire current world annual HALEU consumption of the non-Russian world amounts to perhaps five or six metric tons, some in research reactors and the majority in the Russian icebreaker fleet. Centrus Energy, currently the only Western HALEU enrichment operation, runs a small demonstration cascade in Piketon, Ohio, producing around 900 kilograms per year under a contract with the Department of Energy as its offtaker. One advanced reactor, even a microreactor, can easily require a metric ton or more for its initial fuel loading. The DOE has been effectively subsidizing HALEU supply to backstop the first few advanced reactor demonstrations, but a commercial HALEU fuel cycle requires commercial HALEU demand, and commercial HALEU demand requires a fleet of operating reactors, and building that fleet requires fuel.

This circularity is not peculiar to nuclear. It shows up wherever a new industrial process requires scale to become economical but requires demonstrated economics to justify scale. What makes it particularly acute for advanced nuclear is that the investment required to establish a Western HALEU supply chain is very large, the timeline is very long, and the alternative, light water reactors running on fuel that is already produced in industrial quantity at well-understood cost, is extremely competitive. The startup founders and venture investors now flooding into advanced nuclear may be surprised to learn that the supply chain for their reactor’s fuel essentially does not yet exist in a commercial form. Building it will require sustained demand that no single company can create alone. That challenge is compounded by the sheer number of bespoke reactor vendors in the current landscape, each with their own fuel specification, and in many cases dramatically different fuel types, making any shared infrastructure difficult to justify and harder still to finance robbing them of the economies of scale which help control fuel costs.

The Thread Running Through All of It

What connects the fuel pellet sintering process to the Zero by 2010 initiative to the burnup limits imposed by cladding corrosion to the HALEU supply problem is a single underlying dynamic: nuclear fuel technology improves slowly, through systematic learning from operating experience, and the improvements compound over decades into capabilities that were genuinely not achievable at the start of the program.

The 93 percent capacity factor of the American light water fleet, the near-elimination of fuel failures, the push toward 80 gigawatt-day burnup with LEU Plus, the highly optimized three-dimensional enrichment distributions that extract maximum energy from every kilogram of uranium, none of this was present in the early plants. It accumulated through fifty years of operation, incident investigation, information sharing across the industry, and reinvestment in understanding failure mechanisms. This operationally informed refinement is in itself a technology as important as the nuts and bolts of the reactor itself. The CANDU fuel bundle, which Michael noted has barely changed in its basic geometry across all the reactors in Argentina, Romania, and Canada, represents a different kind of achievement: simplicity so well-matched to its application that it has not required improvement for decades.

The question this history poses for the next generation of reactors is not whether their fuel concepts are interesting, TRISO and metallic fuel both have genuinely attractive properties that I am looking forward to digging into in our next episode. The question is what the equivalent of fifty years of operational learning looks like for technologies that have never operated at commercial scale, and how long it takes before the 93 percent projection that appears in every advanced reactor presentation reflects something other than aspiration.

*This essay accompanies part one of Decouple’s deep dive on the nuclear fuel cycle with Michael Seely. Part two, on TRISO, metallic fuel, and the economics of advanced reactor fuel, will be available soon. Watch on YouTube, or listen on Apple and Spotify.*

*If you find our work valuable, consider supporting Decouple by pledging on Substack or making a tax-deductible donation through our fiscal sponsor at Givebutter.*

Recent conversations about the budding nuclear renaissance often begin with the same list. Microsoft restarting Three Mile Island, tech companies partnering with advanced nuclear reactor vendors or the proposed Japanese funded US AP1000 fleet build. These announcements generate headlines, volatile valuations and investor decks.

Meanwhile, forty Westinghouse pressurized water reactors (PWR) sit at roughly the same thermal power output they were commissioned at decades ago, operating within design margins calculated on transistorized computers before integrated circuits existed. Together, these plants hold between 6-10GW of additional capacity, the equivalent of eighteen to thirty 300MW SMRs, that could be unlocked faster and cheaper than any other nuclear source and perhaps even faster than new gas.

My conversation with returning guests Robb Stewart and James Krellenstein, CTO and CEO of Alva Energy, made the case that power uprates at existing PWR plants represent the lowest-hanging fruit in the sector, bypassing the megaproject risks and nuclear supply chain rebuilding that make new nuclear construction so daunting for utilities.

The components turn out to be far more manageable and anticlimactic than new build nuclear: non-safety-related secondary side equipment such as feedwater heaters and condenser tube bundles, alongside nuclear-grade steam generators that the fleet has already learned to replace during their month long scheduled outages using well-rehearsed industrial choreography.

Alva’s approach avoids the traditional uprate bottleneck by building a separate standardized 250MW Second Turbine Generator Plant (2TGP) building diverting incremental steam from an uprated core without touching the existing turbine centerline. Most of the construction happens on a conventional, non nuclear island using mature supply chains and firm fixed price engineering, procurement, and construction contracts, leaving the outage window limited to a short tie in during a normal refueling cycle.

Compared to a cohort of first of a kind nuclear steam supply system startups that accessed public markets through SPAC mergers and achieved substantial valuations in the hundreds of millions on compelling nuclear narratives this approach sounds deliberately boring. But perhaps boring is what our current moment demands. The question is whether the American nuclear zeitgeist can resist the allure of novelty long enough to pluck the low hanging gigawatts hiding in plain sight.

The uprate asymmetry: BWRs juiced, PWRs stayed natural

The United States nuclear fleet has already added roughly 7GW of capacity through uprates. More than two thirds of that came from boiling water reactors (BWR), which represent only about a third of the operating fleet by unit count. The average BWR in the United States now operates at 160MW above its original licensed thermal power. Some units, particularly in Sweden, have pushed uprates past thirty percent.

Pressurized water reactors, by contrast, have barely moved. A handful of plants have completed small stretch uprates in the 2-4% range. A few larger projects happened in the 1970s and around 2010. The Westinghouse four-loop PWR, which is the single most common nuclear steam supply system in the United States by installed gigawatts, has never completed what the NRC classifies as an extended power uprate above 7%.

This asymmetry is not because PWRs are harder to uprate. The explanation is historical and institutional. General Electric (GE) spent decades refining its understanding of multiphase flow in BWR cores, codifying that knowledge and then systematically carrying those analytical approaches through the regulatory process. By the time they were approved, utilities had a clear, repeatable path to extract more power from the same core geometry.

PWR vendors and utilities, for reasons that had more to do with regulatory caution and shifting business priorities than with physics, never invested in the same systematic effort. The result is a technical arbitrage where core thermal margins and fuel performance allow higher power operation within existing licensing limits, but the industry has lacked a standardized engineering package and industrial execution model to make uprates routine rather than bespoke.

If the opportunity has been sitting there for decades, why did nobody pursue it systematically before? The answer has less to do with engineering than with political economy. For much of the 2000s and 2010s, there was no price signal to reward incremental nuclear output. Load growth was flat or declining in many markets, including deregulated ones where nuclear plants delivered reliability and fuel security but were compensated as low value baseload. Natural gas was cheap, and utilities that maintained nuclear fleets optimized for operations and maintenance cost reduction rather than capital investment to expand output.

Lastly, the project delivery organizations that once executed complex plant modifications had been hollowed out or disbanded, and capital markets placed little value on incremental nuclear megawatts when combined cycle gas plants were faster and cheaper to build and were positioned to capture the highest value tiers in deregulated markets by offering dispatchable capacity and ancillary services.

Share

Why the margin exists: computers, steam generators, conservative assumptions

When Westinghouse and Combustion Engineering designed large PWRs in the late 1960s, computational tools for reactor thermohydraulics were primitive. Safety analyses stacked conservative assumptions because uncertainties could not be resolved with higher fidelity models.

In the interim two things have changed. Operating experience accumulated over thousands of reactor years and modern neutronics codes and thermal hydraulic simulations model core behavior with precision impossible in the slide rule era. A reactor designed in 1968 to very conservative margins can now demonstrate it has been operating well within actual physical limits.

The other major variable is steam generator technology. Early designs were prone to tube fouling, corrosion, and plugging. By the 2000s, the industry mastered steam generator replacement during scheduled outages, often upgrading to designs with better materials and more heat transfer area. A new steam generator with greater thermal capacity pulls more heat from the reactor coolant system without changing anything nuclear.

These dynamics mean many PWRs are constrained not by reactor physics, but by secondary side equipment that was never optimized for higher output and is replaceable during maintenance windows.

Feed water heaters, condensers, and solving the outage bottleneck

The irony of nuclear power uprates is that most of the limiting components are not nuclear grade. Feed water heaters, condenser tube bundles, and main turbine capacity are all balance of plant equipment subject to conventional power generation standards, not 10 CFR 50 Appendix B quality assurance.

This matters because it changes the cost structure and the supply chain. A nuclear grade reactor pressure vessel head replacement is a multi-year procurement involving specialized forging, ASME Section III N-stamp fabrication, and a narrow vendor base. A feed water heater replacement is a large industrial heat exchanger that power plants replace routinely, procured from a much broader supplier ecosystem, installed during a refueling outage with craft labor that does not require nuclear-specific training for most of the work.

However, traditional uprates face a brutal execution problem. Accommodating additional steam from an uprated core often requires upgrading or replacing the main turbine centerline, which can take a plant offline for many months, and in some cases more than a year, while large rotating components are removed, replaced, and recommissioned.

Gigawatt-scale turbine replacements are supply chain nightmares, with seven-year lead times for forgings and machining across global vendors. Refueling outages are already tightly choreographed events with narrow critical paths, and adding major turbine work creates cascading dependencies where a late component delivery or installation complexity can spiral into months of lost generation with average losses of ~1 million dollars per day offline.

The approach being pursued by Alva attempts to sidestep this by standardizing the secondary side uprate package and decoupling it from the outage critical path. Instead of upgrading the existing turbine centerline, the additional steam from the uprated core gets diverted to a new standardized 250MW steam turbine plant, the Second Turbine Generator Plant (2TGP), built outside the protected area while the reactor stays online. During a normal refueling outage, the new plant gets tied into the steam and feedwater headers. To the existing balance of plant equipment, it appears as if the uprate never happened, since the original systems continue operating at their pre-uprate design basis.

Where does the second turbine island actually go? The answer is that many United States sites were originally planned for additional units that were later canceled. Those cancellations left brownfield or greenfield space already served by transmission infrastructure, cooling water systems, and heat rejection capacity.

This configuration changes the economics substantially. The bulk of the work happens on a greenfield site using non-nuclear construction standards, which allows firm fixed-price EPC contracts of the kind that companies routinely execute for combined cycle gas applications. The outage duration drops from months to the time needed for steam generator replacement work that many plants already do in under 30 days. The 200 to 300MW steam turbine supply chain, which has been building these units globally for over a century, is far broader than the narrow vendor base for gigawatt-scale reactor centerline components.

The goal is to build off of examples like Bechtel and SGTe who turned multi-year replacement projects into highly choreographed outages that could swap out a 400-ton steam generator in under thirty days. The learning curve was steep, but once the process was standardized, costs and schedules dropped by an order of magnitude. The first steam generator replacement at Surry in the 1970s took longer than 14 months. By the 2000s, the industry was completing them in 26 days.

The same learning curve logic applies to uprates. The first few projects will be slower and more expensive as engineering margins get validated and installation procedures get refined. By the fifth or sixth unit, the process could become routine, and the economics start to look very different from traditional megaproject math.

Are Uprates Parasitic?

The entire uprate model rests on an assumption about asset longevity. Many of the three-loop and four-loop PWRs being discussed are 1960s and 1970s vintage. Some are well into their second license renewal period. How long can these plants realistically run, and does pushing them to higher thermal power shorten that operating life?

Alva assumes conservatively that uprate economics must clear over a twenty year window, even though many units are likely to operate longer based on subsequent license renewals and ongoing reactor pressure vessel embrittlement monitoring. To date, neither the Nuclear Regulatory Commission nor the utility industry has identified hard material limits that would preclude further extensions, and surveillance reactor vessel capsule data continues to support long term operation.

The asset life question also has a portfolio dimension. The U.S. PWR fleet is not uniform in age. Watts Bar Unit 2 entered service in 2016, Seabrook in 1990, and Vogtle Units 1 and 2 in the early 1990s. These are all Westinghouse four loop reactors, some younger than the engineers now designing uprates for them, spanning an opportunity set that includes plants with decades of remaining life alongside those closer to the end of their initial operating horizon.

Second, and more interesting, is the claim that the uprate configuration can actually improve reliability rather than degrade it. The stress on rotating machinery in a traditional uprate comes from forcing the existing turbines, pumps, and heat exchangers to handle more flow and higher thermal loads. That is where you see unexpected consequences, unplanned trips, thrown turbine blades, and accelerated wear.

By diverting the additional steam to a separate turbine island, the existing balance of plant equipment continues operating at its original design conditions. The reactor coolant pumps run at the same speed. The main turbine sees the same steam flow it always has. In some cases, plants that had been pushing their legacy equipment by a few percent to squeeze out marginal gains can now dial that back, sending the extra steam to the new turbine and recovering reliability margin on the old equipment.

The nuclear steam supply system itself, according to this argument, is not being stressed in the conventional sense. The new steam generators are designed for the higher thermal duty. The reactor core operates at a higher power level, but modern fuel can handle it, and neutron flux increases are accounted for in embrittlement calculations. In some configurations, the primary system actually runs slightly cooler than before because the improved heat transfer pulls more energy out at lower temperatures.

Steam generator replacement has historically been life-extending rather than life-consuming. If the uprate layers additional revenue on top of that investment without imposing new mechanical stress on the equipment that actually fails, the model works.

Share Decouple

What actually drives the schedule?

A claimed five-year timeline from engineering to commercial operation does not immediately sound transformational. Combined cycle gas plants can go from greenfield to energization faster than that in many jurisdictions, and the pitch is that uprates could potentially beat gas on speed. What takes so long?

The answer is procurement, specifically steam generator fabrication. There has not been a robust market for new steam generators in recent years, and the vendor base needs time to scale back up. Unlike balance of plant components, steam generators are nuclear-grade pressure vessels subject to ASME Section III requirements, and the forging and tube bundle fabrication cannot be rushed. That procurement dominates the critical path for the first project.

The critical difference from traditional uprates is what is *not* on the critical path. All of the balance of plant construction, the turbine island fabrication, the installation of condensers and feedwater heaters, happens in parallel in the Second Turbine Generator Plant while the reactor stays online. By the time the new steam generators are ready and the plant goes into a refueling outage, the 2TGP is already built and tested. The outage itself only needs to accommodate the steam generator swap and the tie-in to the new headers, work that fits within a normal refueling window that the plant would be taking anyway.

This architecture allows parallelization. Once the first project validates the design and establishes the supply chain, subsequent projects at other sites can proceed simultaneously. The developer is not waiting for one plant to finish before starting the next. Multiple steam generator sets can be in fabrication for different sites. Multiple turbine islands can be under construction. The bottleneck shifts from sequential field construction to fabrication capacity, and fabrication scales faster than field work.

There is also an option value embedded in the model. If a hyperscaler wants power faster and is willing to pay for acceleration, the plant can take an earlier outage, compressing the overall schedule at the cost of higher near-term spending. That flexibility does not exist in new build projects, where the critical path is locked in by the construction sequence.

Whether five years is realistic for the first project is testable. Whether the timeline compresses for subsequent units depends on how well the standardization actually works and whether the steam generator supply chain can ramp without hitting quality or delivery problems. The schedule claim is aggressive but not obviously implausible, and it is far more credible than any of the timelines being quoted for first-of-a-kind reactor designs.

The Microsoft arbitrage: $116/MWh for 800 megawatts

The demand signal for additional nuclear capacity is no longer theoretical. Microsoft signed a power purchase agreement with Constellation to restart Three Mile Island Unit 1 at a reported price of $116/MWh. That is multiples of the current PJM wholesale clearing price, and it is for 835MW of capacity that was already built, already licensed, and sitting idle.

This price point changes the economics of everything. A nuclear uprate that adds 200-300MW to an existing unit, if it can secure even a fraction of that premium through a long-term PPA with a hyperscaler or large industrial customer, suddenly has a revenue stream that supports very aggressive capital investment.

The reason those prices exist is simple. Data centers need firm, 24/7 power with very high reliability, and although climate policy is on the back burner some still value low-carbon generation to satisfy their corporate commitments and comply with their AI customers’ environmental procurement requirements which is why Constellation can command that kind of premium.

This dynamic is not limited to restarts. An uprate that brings 200-300MW of additional capacity online at an existing site, with an existing transmission interconnection and an operating license already in hand, competes not against wholesale power markets but against the all-in cost of new generation that can meet the same reliability and carbon requirements. In that comparison, uprates start to look very attractive indeed.

The time advantage is also real. A well-executed uprate can go from engineering to commercial operation in 3-5 years, depending on the licensing path and the outage schedule. That is faster than any new nuclear build and faster than most large combined cycle gas plants in contested permitting environments.

If the hyperscalers are serious about expanding their AI infrastructure over the next decade, and if they are willing to pay nuclear premiums to secure that expansion, then there are billions of dollars of potential offtake value sitting in the existing PWR fleet waiting to be unlocked.

A typical 20 percent uprate on a 1,200MW unit adds 240MW, almost the equivalent of a BWRX-300 or roughly three NuScale power modules, at one-fifth to one-seventh the cost and without the construction risk or regulatory complexity of first-of-a-kind designs. Multi-unit stations could add several Alva Second Turbine Generator Plants (2TGPs) the equivalent of multiple SMRs. The entire uprate potential across the PWR fleet represents 20 to 30 small modular reactors worth of capacity, available in three to five years rather than fifteen.

The hidden crux: Competent project delivery organizations

One of the most common objections to nuclear uprates is regulatory risk. The reality is that the NRC has a well-established framework for reviewing power uprates, and utilities have successfully completed dozens of them under both the standardized review process for power uprates and through the license amendment process.

The real question is whether the American nuclear industry can rebuild the kind of project delivery organizations that used to exist at utilities like Florida Power & Light (FPL). The difference between a good nuclear project and a disaster often comes down to the strength of the owner or developer organization, not the reactor vendor. This is the lesson from comparing the construction of nuclear power plants St. Lucie Unit 2 in Florida with Washington Nuclear Project 3 in Washington State. Both were Combustion Engineering reactors with Ebasco as architect engineer, both under the NRC, both starting construction in 1977.

St. Lucie 2 came in under budget in about five years in the post-Three Mile Island context. It was built by FPL working with a couple hundred person internal projects organization whose only job was to deliver new nuclear for the utility.

In sharp contrast, Washington Nuclear Project 3 helped trigger the largest municipal bond default in American history at the time, built by the Washington Public Power Supply System, a loose consortium of dozens of small public utilities that had never built large thermal plants before.

The timing, hardware, vendor, and regulator were identical, yet outcomes diverged because the owner side organization differed in its ability to integrate the work, supervise execution, and enforce discipline.

A project delivery organization is not a contractor. It is an entity that takes responsibility for integrating design, procurement, construction, commissioning, and handover, and that has the authority and accountability to enforce schedule and cost discipline across all of those phases. It sits between the utility that will own and operate the asset and the vendors that supply the hardware. It knows how to read a Primavera P6 project schedule, manage a critical path, run value engineering sessions, and make the hard calls about when to push and when to stop.

This kind of organization does not get built by raising venture capital or hiring consultants. It gets built by doing hard projects, making mistakes, learning from them, and keeping the team together long enough to apply those lessons to the next project.

In contemporary terms Canadian utilities, Ontario Power Generation (OPG) and Bruce Power, did not suddenly know how to run billion-dollar refurbishments. They learned by doing them, repeatedly, under enormous public scrutiny, and they kept the institutional memory alive by maintaining continuity in leadership and core staff. Ultimately both OPG and Bruce Power have delivered under budget and ahead of schedule.

The strategy of starting with uprates rather than jumping straight to new builds offers a sequencing that allows organizations to build capability progressively. If a developer can repeatedly bring 200-300MW increments online at multiple sites over the next five years, they will have demonstrated something that almost no one else in the American nuclear sector can claim right now.

Boring is the new radical

The nuclear industry has spent the last two decades chasing innovation in advanced reactors, modular designs and factory fabrication with very little to show for it.

Power uprates at existing PWRs represent industrial execution and process discipline rather than the kind of work that generates TED talks or 500 million dollar SPAC combinations, but they actually put gigawatts on the grid. At this point if the nuclear renaissance is going to be real, it will not be because someone invented a better reactor. It will be because someone figured out how to repeatedly deliver projects on time and on budget, and then scaled that capability across a fleet.

The hardware already exists, the regulatory path is clear, and the demand is there, but what has been missing is the organizational capacity to turn all of that into megawatts. Time will tell whether developers like Alva can execute on the first few projects, prove the model works across different sites and vintages, and build the kind of repeatable project delivery organization that can scale beyond uprates. If they can, the nuclear sector gains something it has lacked for decades: a credible path from concept to commissioning that does not depend on heroic assumptions or government guarantees.

The most boring path to ten new gigawatts might also be the best path, for now.

If this analysis informed your thinking, consider supporting independent nuclear journalism. Make a tax-deductible donation

Discussion about this essay is open to all subscribers. If you have thoughts on uprates, project delivery, or the broader nuclear landscape, share them in the comments.