On April 15, 2023, Germany switched off its last three nuclear reactors \u2014 Isar 2, Emsland, Neckarwestheim 2 \u2014 while its gas-fired fleet kept running. The lights didn’t flicker. The grid held. Nobody outside the energy press noticed much. And that was the point, or at least the political bet: that the absence of nuclear would be invisible, absorbed by other sources, forgotten.<\/p>\n\n\n\n
A PricewaterhouseCoopers counterfactual analysis, commissioned by Strategy& and reported by NucNet in January 2025, modeled what would have happened had Germany retained the seventeen nuclear reactors it had operational in 2010. The answer: 94% of its 2024 electricity generation would have been emission-free. That’s modeling, not measurement. But the measured reality isn’t in dispute either. Fossil gas accounted for roughly 16% of German generation in 2024. Coal still sat above 20%. The government has announced plans for 20 GW of new gas-fired capacity \u2014 plants that will burn fossil fuel for decades. A country that talks about climate more earnestly than almost any other in Europe deliberately removed zero-carbon generation and filled the gap with combustion. Not by accident. Not by oversight. By policy choice, sustained over two decades, executed on schedule, and celebrated as a moral accomplishment.<\/p>\n\n\n\n
Germany is not exceptional. It is the pattern.<\/p>\n\n\n\n
Mention nuclear power and people think of catastrophe \u2014 Chernobyl, Fukushima, the mushroom cloud that nuclear energy has never managed to fully separate itself from. Or they think about the future: small modular reactors, fusion, technologies that might eventually be built if the funding holds and the physics cooperates. What almost nobody thinks about is the technology that already exists, already works, and already produces more low-carbon electricity than any single source except hydropower.<\/p>\n\n\n\n
The global nuclear fleet \u2014 413 operational reactors in 31 countries \u2014 generated a record 2,667 TWh of electricity in 2024, surpassing the previous high of 2,660 TWh set in 2006. That’s more than wind produces globally. More than solar. It represents roughly 9% of the world’s electricity \u2014 the lowest share in four decades, down from 17.5% in 1996, even as absolute output just hit an all-time peak. The numbers tell two stories at once: nuclear is enormous, and nuclear is shrinking. The shrinking share isn’t because nuclear is generating less \u2014 it’s because global electricity demand has more than doubled since 1996, and almost all that growth was met by gas, coal, and more recently renewables. Nuclear stayed roughly flat. The world grew around it.<\/p>\n\n\n\n
Within low-carbon generation specifically, nuclear trails only hydropower at 14.3% and exceeds both wind at 8.1% and solar at 6.9%, according to Ember’s 2025 Global Electricity Review. The US fleet runs at a capacity factor of 92% \u2014 meaning plants generate electricity 92% of all hours in the year, outperforming every other energy source. Globally, the average is 83%. These plants don’t wait for wind. They don’t stop when the sun goes down. They run, continuously, for eighteen months between refueling outages, producing zero-carbon electricity around the clock.<\/p>\n\n\n\n
The fleet is old. More than two-thirds of operational capacity \u2014 295 reactors representing 261.8 GWe \u2014 has run for over 30 years. The average age in advanced economies is 36 years. But age hasn’t degraded performance. IAEA data shows reactors exceeding 45 years of operation maintain higher-than-average capacity factors. The oldest plants are, paradoxically, among the most reliable. France’s 56 reactors give it one of the lowest-carbon electricity grids in Europe \u2014 per-kilowatt-hour emissions a fraction of Germany’s despite comparable industrial output. The comparison is not subtle: France kept its nuclear fleet and has clean electricity. Germany closed its fleet and burns gas.<\/p>\n\n\n\n
This isn’t hypothetical clean energy. It’s actual clean energy, operating at scale, connected to grids, staffed and running right now. And the thing that threatens it isn’t radiation or earthquake risk or the problem of waste storage.<\/p>\n\n\n\n
It’s a spreadsheet.<\/p>\n\n\n\n
In liberalized electricity markets \u2014 the kind that exist across most of Europe, much of the United States, and increasingly worldwide \u2014 generators get paid the wholesale spot price. That price is set by marginal cost: the cost of the most expensive plant needed to meet demand at any given moment. Nuclear plants have very high capital costs, already sunk. They have very low fuel costs \u2014 uranium is cheap. But they carry significant fixed operating costs that cannot be avoided: staffing (a large plant employs 500\u2013700 people), maintenance, regulatory compliance, waste handling, security. These costs don’t change whether the plant runs or not. They’re the price of being available.<\/p>\n\n\n\n
When wholesale prices fall \u2014 driven down by cheap natural gas and zero-marginal-cost renewables flooding the market during peak hours \u2014 nuclear’s revenue drops below what it costs to keep the doors open. The plant is profitable to run for one more hour; its marginal cost is among the lowest on the grid. But it is unprofitable to keep open for one more year, because its fixed costs exceed its total market revenue over that period.<\/p>\n\n\n\n
“Uneconomic” doesn’t mean the electricity is too expensive to produce. It means the market won’t pay enough to cover the cost of being there to produce it.<\/p>\n\n\n\n
The Union of Concerned Scientists quantified this in 2018. Their report “The Nuclear Power Dilemma” found that more than a third of US nuclear plants \u2014 21 of 60 facilities \u2014 were unprofitable or scheduled for closure. The annual shortfall across the at-risk fleet: roughly $814 million. Not a staggering number in the context of the US energy system. But enough, in a deregulated market with no mechanism for compensating zero-carbon baseload, to kill plant after plant.<\/p>\n\n\n\n
Since 2012, twelve US reactors have permanently closed, representing approximately 9,400 MW of capacity \u2014 nearly 10% of the American nuclear fleet. Vermont Yankee closed in 2014. Pilgrim in 2019. Indian Point in 2021. In each case, the operator \u2014 Entergy, in all three \u2014 cited low natural gas prices and rising operating costs. These plants received 20-year license renewals from the NRC just three to five years before their owners announced closure. The NRC \u2014 the safety regulator \u2014 said they were safe to operate for decades more. The market \u2014 the economic regulator, in effect \u2014 said they weren’t worth keeping. The two judgments have nothing to do with each other, and that disconnect is the entire problem.<\/p>\n\n\n\n
And the mechanism doesn’t stop with past closures. In Spain, the government’s plan to end all nuclear by 2035 puts seven operational reactors on a political countdown. Almaraz \u2014 2.0 GW, Spain’s largest \u2014 goes first: Unit 1 in November 2027, Unit 2 in October 2028. Its operators have requested a three-year extension to mid-2030. The government has set conditions that amount to refusal \u2014 no cost to taxpayers, guaranteed safety, conditions that no other generator in the country faces. In the US, the Section 45U production tax credit created by the Inflation Reduction Act expires in 2032. Without renewal, the entire American fleet re-enters the economic exposure that closed a dozen reactors before it. The European Commission warned in March 2026 against premature closure of nuclear plants. It acknowledged the pattern. It lacks the authority to prevent it.<\/p>\n\n\n\n
How a cheap-to-run plant loses money: the merit order<\/strong>\n\nElectricity markets dispatch generators in order of marginal cost \u2014 cheapest first. Wind and solar have a marginal cost of effectively zero: once built, the fuel is free. When they generate, they push the market clearing price down. Gas plants typically set the price when renewables aren't generating. Nuclear sits in between: its marginal cost is low (fuel is cheap), but its fixed costs are high (staff, maintenance, compliance, security). When cheap gas and zero-cost renewables push wholesale prices below nuclear's total operating cost for enough hours of the year, the plant loses money \u2014 even though every megawatt-hour it produces costs almost nothing at the margin. The market doesn't pay generators for being available. It pays for the next unit of energy. That's a design choice, not a law of nature. And it's a design choice that systematically undervalues any generator whose primary benefit \u2014 always on, zero-carbon \u2014 doesn't show up in the spot price.<\/code><\/pre>\n\n\n\nThe replacement problem<\/h3>\n\n\n\n
When a nuclear plant closes, something has to generate the electricity it was producing. Wind and solar can’t replicate what nuclear provided \u2014 their output depends on weather, and the grid needed nuclear’s round-the-clock, rain-or-shine reliability. What fills the gap, documented case after case, is fossil fuel.<\/p>\n\n\n\n
Indian Point, New York. Two gigawatts of capacity. Closed April 30, 2021 \u2014 Unit 2 had already shut down a year earlier. In the first full month after the final closure, natural gas generation’s share of New York’s electricity supply jumped from 35% to 39%. Three gas-fired plants were built or expanded in the region to fill the void: Bayonne Energy Center, CPV Valley Energy Center, Cricket Valley Energy Center \u2014 combined capacity of 1.8 GW, all burning fossil gas, replacing 90% of Indian Point’s nameplate capacity with combustion. The Foundation for Research on Equal Opportunity calculated that in 2022, Indian Point’s absence produced an additional 8.03 million metric tons of CO2 that would not have been emitted had the plant kept running. For scale: that’s more than the entire annual emissions of Costa Rica, a country of 5.2 million people.<\/p>\n\n\n\n
One plant. One closure decision. Eight million tons of CO2.<\/p>\n\n\n\n
The pattern holds across the country. Luke Petach’s 2025 study in the Southern Economic Journal \u2014 rigorous, peer-reviewed, using difference-in-differences analysis across every US nuclear closure from 1993 to 2022 \u2014 found that each nuclear plant closure increases state-level per capita carbon emissions by 6\u20138%. The driver: substitution toward fossil fuel generation. In earlier closures, that meant coal. In more recent ones, increasingly gas. E&E News documented the same dynamic in Massachusetts, New York, and Pennsylvania after reactor shutdowns: gas generation filling the void each time. In every documented case \u2014 every closure the data covers \u2014 fossil fuels filled the gap, not renewables.<\/p>\n\n\n\n
Germany offers the largest-scale version of the same story. The PwC counterfactual models 94% emission-free generation had the nuclear fleet been retained. The measured reality, as of 2024: fossil gas at 16%, coal above 20%, and a government planning to build 20 GW of new gas plants. The gap between what Germany’s grid could have looked like and what it actually looks like is the most expensive climate mistake in European energy history \u2014 made not once, in a moment of panic after Fukushima, but repeatedly, over more than a decade, in the full knowledge of what the emissions arithmetic would produce.<\/p>\n\n\n\n
The damage goes beyond carbon. Freese et al., in a 2023 study in Nature Energy, modeled the air quality consequences of nuclear retirement across the US. Replacing nuclear with fossil generation produces measurable increases in fine particulate matter and ground-level ozone. Their estimate: 5,200 additional premature deaths annually from the air quality degradation alone \u2014 people who die of respiratory and cardiovascular disease because a zero-emission plant was replaced by one that burns fuel. Climate damages from a single year of the replacement emissions: $11\u2013180 billion.<\/p>\n\n\n\n
The rescues and their limits<\/h3>\n\n\n\n
Some plants have been saved. The mechanisms are instructive \u2014 and their fragility is the point.<\/p>\n\n\n\n
New York pioneered Zero-Emission Credits in 2016, requiring utilities to purchase credits from in-state nuclear plants as compensation for their carbon-free attribute. The program saved Nine Mile Point and Fitzpatrick from closure. Exelon purchased Fitzpatrick from Entergy specifically because ZECs made continued operation financially viable \u2014 without the credits, the economics didn’t work. Extended through 2049 in January 2026 by the state Public Service Commission, these four upstate reactors now provide more than 40% of New York’s zero-emission power. The irony is stark: New York created a program to save nuclear plants upstate while simultaneously celebrating the closure of Indian Point downstate. The emissions data suggests it saved the wrong ones \u2014 or rather, failed to save all of them.<\/p>\n\n\n\n
Illinois adopted a similar model. At the federal level, the Inflation Reduction Act’s Section 45U production tax credit offers up to 1.5\u00a2\/kWh for existing nuclear plants through 2032 \u2014 designed explicitly to prevent economic closures. In early 2025, these credits accounted for roughly 25% of all transferable clean energy tax credit transactions, priced at $0.96\u20130.97 per dollar of credit value \u2014 among the highest in any category, reflecting strong market confidence that the credits would be used.<\/p>\n\n\n\n
Diablo Canyon, California’s last nuclear plant, was scheduled for closure in 2024\u20132025 under a 2016 agreement between PG&E and environmental groups. Governor Newsom signed SB 846 in September 2022 reversing the decision. The NRC approved a 20-year license renewal on April 2, 2026. Across the Atlantic, Belgium’s parliament voted 102\u20138 on May 15, 2025, to repeal its 2003 nuclear phase-out law entirely. Energy Minister Bihet outlined a target of 8 GW of nuclear capacity by 2035 \u2014 retaining existing plants through life extensions and adding new build. The vote wasn’t close. The previous policy, it turned out, had virtually no defenders left.<\/p>\n\n\n\n
The political winds have shifted enough that in May 2025, the Trump administration issued executive orders setting a goal of US “global leadership” in nuclear energy, including 5 GW of power uprates to existing reactors \u2014 squeezing more output from plants already built.<\/p>\n\n\n\n
But every intervention listed here is reactive, expensive, plant-by-plant, and politically contingent. No country has built a systematic market mechanism for valuing what nuclear provides \u2014 reliability, zero-carbon baseload, fuel security \u2014 within the structure of its wholesale electricity market. Every rescue requires a separate political fight, a separate legislative act, a separate subsidy mechanism. The \u00a745U credit expires in 2032 with no guarantee of renewal. Executive orders last as long as the administration that signs them. Plants whose states lack a political champion willing to spend capital on the fight simply close. And once they close, the option to reverse that decision is, for practical purposes, gone.<\/p>\n\n\n\n
What \"we can always restart it later\" actually costs<\/strong>\n\nPalisades, Michigan: closed May 2022. It is attempting to become the first US commercial reactor to restart after entering decommissioning. The price tag: $1.52 billion in DOE loan guarantees plus $150 million from the state of Michigan. The restart timeline has already slipped to 2026 due to steam generator tube sleeving. Compare this to what would have prevented the closure in the first place: an annual operating subsidy in the range of $50\u2013100 million \u2014 a fraction of the restart cost. Once a plant shuts down, the specialized workforce disperses, fuel is removed, systems degrade, regulatory status changes. Restarting is not flipping a switch. It is a billion-dollar, multi-year construction project requiring fresh NRC approvals at every stage. For the overwhelming majority of closed reactors, restart will never be attempted. Closure is, in practice, permanent. The reassuring fiction that we can shut plants now and reopen them later costs nothing to believe and billions to disprove.<\/code><\/pre>\n\n\n\nWhy nuclear cannot make its own case<\/h3>\n\n\n\n
A question hangs over all of this: why does a problem this obvious, this well-documented, this quantifiable, persist? The economic mechanism is understood. The emissions consequences are measured to the megatonne. Policy instruments exist and have been deployed successfully in specific cases. And yet the political response remains piecemeal, reactive, perpetually behind the closures it’s trying to prevent.<\/p>\n\n\n\n
Start with the industry itself. The nuclear sector’s default public posture is defensive \u2014 responding to safety fears with data about safety, which concedes the frame that safety is the relevant question for existing plants. When someone asks whether a nuclear plant should stay open, the industry answers with probability-of-core-damage-per-reactor-year. The public hears: they’re talking about damage. The actual argument \u2014 that the operating nuclear fleet is the largest-scale source of zero-carbon electricity humans possess, and that closing it while burning gas is climate vandalism \u2014 has been made almost exclusively by outsiders: climate scientists like James Hansen, energy modelers, heterodox environmentalists who broke with their organizations over the issue. Not by the industry. The Nuclear Energy Institute’s public communications overwhelmingly emphasize safety records and operational reliability rather than climate urgency or what nuclear saves ratepayers compared to the fossil alternative. An industry that cannot articulate why it matters in terms that reach beyond people already converted has, in a real sense, chosen to lose.<\/p>\n\n\n\n
Then there’s the organized opposition. The Sierra Club states its position plainly: “unequivocal” opposition to nuclear energy. Not conditional opposition. Not opposition to new plants while supporting existing ones. Unequivocal. NRDC senior scientist Mohit Chhabra wrote in June 2021 that the organization had “been working for years to retire the aging Diablo Canyon nuclear power plant.” NRDC celebrated the closure of Indian Point that same year. In both Indian Point and Diablo Canyon, the direct consequence of the positions these organizations fought for \u2014 and in Indian Point’s case, won \u2014 was more natural gas combustion and higher carbon emissions. Outcomes they nominally oppose. The contradiction isn’t hidden. It simply doesn’t cost them anything politically to maintain it.<\/p>\n\n\n\n
Only in March 2026 did NRDC crack. It filed comments with the NRC supporting the Duane Arnold reactor restart in Iowa \u2014 the first time in the organization’s history it had supported any nuclear plant. President Manish Bapna acknowledged the milestone. The filing came after coal generation in Iowa rose sharply between 2024 and 2025, making the emissions cost of the anti-nuclear position visible enough to be untenable. Fifty years of blanket institutional opposition. Then one narrow, plant-specific, heavily caveated reversal \u2014 and only after the evidence in one state became impossible to look away from.<\/p>\n\n\n\n
The political asymmetry underneath all this is structural. Closing a nuclear plant satisfies an organized, funded environmental constituency and imposes no visible cost \u2014 the emissions increase is diffuse, delayed, attributable only through grid data that nobody outside the energy sector reads. Keeping a plant open satisfies no organized constituency. Ratepayers don’t march for baseload. It requires a governor or a legislator to spend political capital fighting opponents who will name them publicly, run campaigns against them, frame the issue as a safety question. The concentrated political costs of keeping nuclear plants open versus the diffuse, invisible climate costs of closing them \u2014 that asymmetry is why governments act against their own stated decarbonization targets. The emissions math is public \u2014 these governments publish it. The political math points the other direction.<\/p>\n\n\n\n
It’s not ignorance. It’s incentives.<\/p>\n\n\n\n
The fleet is aging. The average reactor in advanced economies is 36. License extensions buy time \u2014 twenty years, when they’re granted. But time is what climate change doesn’t offer. And the political machinery for using that bought time \u2014 for converting a 20-year license extension into 20 years of zero-carbon generation rather than a slow march toward politically convenient closure \u2014 barely exists anywhere in the world.<\/p>\n\n\n\n
Germany closed its nuclear fleet and now burns gas where it didn’t need to. The answer to how that happened is no longer mysterious: market design that can’t price what nuclear provides, organized opposition from the very groups claiming to fight climate change, an industry that brings probability statistics to a political fight, and an incentive structure that rewards closure and punishes defense. The pattern is legible. And it’s about to repeat.<\/p>\n\n\n\n
Almaraz sits on the dry plains of Extremadura in western Spain. Two gigawatts. Spain’s largest nuclear plant. It generates roughly 7% of the country’s electricity and powers four million homes. Unit 1 is scheduled for closure November 1, 2027. Unit 2, October 31, 2028. The closure has nothing to do with the reactors’ condition \u2014 they could run for years more. It is purely a political timetable, set by a government that has promised to end all Spanish nuclear by 2035. After Almaraz: Asc\u00f3, Vandell\u00f3s, Cofrentes, Trillo \u2014 all scheduled to go dark within the decade.<\/p>\n\n\n\n
Iberdrola, Endesa, and Naturgy have formally requested a three-year extension to mid-2030. The European Parliament has urged Spain to reconsider. The Spanish government has set conditions amounting to refusal.<\/p>\n\n\n\n
As of this writing, Almaraz is still running. Still producing zero-carbon baseload. Still invisible to anyone not reading the grid data. Its closure date was set by people who will never have to account for the emissions that replace it.<\/p>\n\n\n\n