The Atom Returns: Why the World's Most Feared Energy Source Is Its Best Hope

In September 2024, Microsoft signed a twenty-year power purchase agreement to restart Unit 1 of the Three Mile Island nuclear plant in Pennsylvania — the very facility whose partial meltdown in 1979 became the defining image of nuclear danger in the American imagination. The deal, worth an estimated $16 billion over its lifetime, was not driven by nostalgia or contrarianism. It was driven by electricity. Microsoft’s AI data centres consume staggering quantities of power — a single large facility can draw 100 megawatts or more, continuously, around the clock — and the company had concluded that neither solar panels nor wind turbines could guarantee the uninterrupted baseload supply that machine learning demands (Constellation Energy, 2024). Google followed weeks later with a deal to purchase power from Kairos Power’s advanced small modular reactor. Amazon signed agreements with Talen Energy for nuclear-adjacent data centre capacity in Pennsylvania. In the space of a few months, three of the most powerful technology companies on earth placed enormous bets on a technology that polite opinion had spent forty years declaring dead.

The irony is rather thick. The same Silicon Valley that lectures the world on sustainability has discovered that when you need power that actually shows up at 2 a.m. on a windless January night, you need something more reliable than good intentions and a solar farm. The Germans, who spent the better part of €500 billion on their Energiewende — the most expensive energy transition in human history — only to shut down their last three nuclear plants in April 2023 and then fire up coal and imported Russian gas to keep the lights on, might have offered a cautionary word. The French, who generate roughly 70 per cent of their electricity from nuclear power at a carbon intensity less than half of Germany’s, were too busy drinking inexpensive wine under affordable electric lighting to comment (Ember, 2024).

The atom, it turns out, never really left. It merely waited for the world to exhaust every fashionable alternative before knocking on the door again.

Atoms for Peace — The First Nuclear Age

The civilian nuclear age began, as so many consequential things do, with a speech. On 8 December 1953, President Dwight D. Eisenhower stood before the United Nations General Assembly and delivered what became known as the “Atoms for Peace” address. The Cold War was at its most frigid. The Soviet Union had detonated its first hydrogen bomb four months earlier. Eisenhower, a military man who understood the implications better than most, proposed that the superpowers redirect fissile material from weapons stockpiles toward peaceful energy generation. “The United States pledges before you,” he declared, “its determination to help solve the fearful atomic dilemma — to devote its entire heart and mind to find the way by which the miraculous inventiveness of man shall not be dedicated to his death, but consecrated to his life” (Eisenhower, 1953).

It was magnificent rhetoric. More unusually, it was followed by action. The International Atomic Energy Agency was established in 1957. The United States launched its first commercial nuclear reactor at Shippingport, Pennsylvania, in 1958. Britain, which had already opened the world’s first industrial-scale nuclear power station at Calder Hall in 1956, pressed ahead with its Magnox reactor programme. The Soviet Union, not to be outdone, had connected the Obninsk reactor to the grid in 1954, technically making it the first nuclear plant to generate electricity for civilian use. The atom was democratising faster than anyone had expected.

But the nation that embraced nuclear power most decisively — and most consequentially — was France. The story begins with the 1973 oil crisis. When OPEC quadrupled the price of crude following the Yom Kippur War, the shock was felt nowhere more acutely than in France, which imported virtually all of its oil and had almost no domestic fossil fuel reserves. Prime Minister Pierre Messmer, acting on plans that had been germinating since de Gaulle’s era, announced what became the Plan Messmer: France would build enough nuclear reactors to achieve near-total energy independence from Middle Eastern oil. It was an audacious bet. Over the next fifteen years, France constructed 56 pressurised water reactors — a rate of roughly one every three months. By the mid-1980s, nuclear provided over 70 per cent of French electricity, a share it has maintained, with modest variation, ever since (World Nuclear Association, 2024).

The French approach was, characteristically, both centralised and pragmatic. Electricité de France, a state-owned monopoly, standardised on a single reactor design — the Westinghouse-derived pressurised water reactor — and built them in series, driving down costs through repetition and institutional learning. The contrast with Britain, which experimented with multiple reactor types (Magnox, Advanced Gas-Cooled Reactors, then switched to pressurised water reactors), and the United States, where every reactor was essentially a bespoke construction project subject to endless regulatory revisions, could not have been sharper. France proved that nuclear could be built quickly, cheaply, and at scale — if, and only if, a government committed fully and did not change its mind every electoral cycle.

Across the Atlantic, the United States built the world’s largest nuclear fleet — 104 reactors at its peak — but did so in a characteristically American fashion: expensively, litigiously, and with mounting public anxiety after Three Mile Island in 1979. By 1990, nuclear power supplied approximately 17 per cent of global electricity, with 413 operational reactors across 31 countries (Smil, 2017). The atom had delivered on Eisenhower’s promise. Then something changed.

The Fear Years — Chernobyl, Fukushima, and the Green Backlash

At 1:23 a.m. on 26 April 1986, Reactor Number Four at the Chernobyl Nuclear Power Plant in Soviet Ukraine suffered a catastrophic steam explosion during a botched safety test. The resulting fire burned for ten days, releasing a plume of radioactive material that drifted across Europe, contaminating farmland in Scandinavia, triggering iodine tablet panic in Germany, and depositing caesium-137 in Welsh sheep pastures that remained under restriction until 2012. Chernobyl killed 31 people directly and caused an estimated 4,000 to 16,000 excess cancer deaths over subsequent decades, depending on the dose-response model applied (UNSCEAR, 2008). It also killed something less measurable: public trust in nuclear energy.

The psychological impact was disproportionate to the physical toll — and deliberately so. Spencer Weart, in his magisterial study The Rise of Nuclear Fear (2012), traces how nuclear energy was never evaluated on its engineering merits alone. From the beginning, the atom was entangled with apocalyptic imagery — Hiroshima, fallout shelters, mutant monsters, the end of civilisation. Chernobyl confirmed every fear that popular culture had been rehearsing since the 1950s. It did not matter that the RBMK reactor design that exploded at Chernobyl was a Soviet military derivative with no containment building, a positive void coefficient that made it inherently unstable at low power, and an operating culture that treated safety protocols as suggestions. What mattered was the image: a burning reactor, an invisible poison cloud, and evacuated cities. The image was indelible.

The anti-nuclear movement, which had been growing since the 1970s, seized the moment. In Germany, the Greens — founded in 1980 with nuclear opposition as a founding principle — gained political momentum. Italy held a referendum in 1987 and voted to phase out nuclear power entirely. Sweden announced a phase-out (later reversed). Austria, which had built a reactor at Zwentendorf but never switched it on after a 1978 referendum, congratulated itself on its foresight. Across Western Europe, nuclear became politically toxic.

Then came Fukushima. On 11 March 2011, a magnitude 9.0 earthquake off the coast of Japan triggered a tsunami that overwhelmed the seawalls at the Fukushima Daiichi Nuclear Power Plant. Three reactors suffered meltdowns. The direct death toll from radiation was zero — not a single person has died from radiation exposure attributable to Fukushima, according to the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR, 2021). The earthquake and tsunami killed approximately 20,000 people. But it was the reactors, not the wave, that dominated global headlines.

Germany’s response was immediate and dramatic. Chancellor Angela Merkel, a physicist by training who might have been expected to evaluate the evidence dispassionately, ordered the permanent shutdown of Germany’s oldest reactors within days and announced a full nuclear phase-out by 2022. Italy held another referendum and reconfirmed its anti-nuclear stance. Switzerland announced a phase-out. Japan shut down its entire fleet of 54 reactors, most of which remained offline for years.

The paradox is striking. Nuclear energy is, by every available statistical measure, the safest form of electricity generation per unit of energy produced. Our World in Data, drawing on peer-reviewed mortality estimates, calculates that nuclear causes 0.03 deaths per terawatt-hour of electricity — compared with 24.6 for coal, 18.4 for oil, and 2.8 for natural gas (Ritchie, 2020). Even accounting for Chernobyl and Fukushima, nuclear kills fewer people than rooftop solar installations (which involve falls from heights) and vastly fewer than the air pollution generated by the fossil fuels that replace it when reactors close. Germany’s nuclear exit is estimated to have caused approximately 1,100 additional deaths per year from increased air pollution as coal plants ramped up to fill the gap (Jarvis, Deschenes and Jha, 2022).

The historical parallel is instructive. When the Luddites smashed textile machinery in the English Midlands between 1811 and 1816, they were not irrational. They correctly perceived that mechanisation would destroy their livelihoods. But the machinery was not the enemy — the absence of transition support was. The anti-nuclear movement made a similar categorical error. It correctly identified that nuclear technology carries risks. It incorrectly concluded that the risks were worse than the alternative — which, in practice, was not a pastoral utopia of wind and solar, but the continued burning of fossil fuels at industrial scale.

The Renaissance — Why Everything Changed

Something shifted around 2022. The factors that had kept nuclear in political exile — public fear, green opposition, cheap natural gas, and the falling cost of renewables — either reversed or proved insufficient. What emerged was not a single cause but a convergence of five forces, each reinforcing the others.

First, climate arithmetic. The Intergovernmental Panel on Climate Change stated plainly in its Sixth Assessment Report that most pathways to limiting warming to 1.5°C require a significant expansion of nuclear power (IPCC, 2022). The International Energy Agency’s Net Zero by 2050 roadmap calls for a doubling of nuclear capacity, noting that “nuclear is the second-largest source of low-carbon electricity today” and that achieving net zero without it would be “more difficult and more expensive” (IEA, 2021). The arithmetic is unforgiving: solar and wind are intermittent, battery storage remains prohibitively expensive at grid scale for multi-day events, and no major industrial economy has decarbonised its electricity supply without either nuclear or exceptional hydroelectric endowments. France’s grid emits roughly 56 grams of CO₂ per kilowatt-hour. Germany’s emits approximately 385 grams — nearly seven times as much — despite Germany having spent far more on renewables (Ember, 2024).

Second, energy security. Russia’s invasion of Ukraine in February 2022 exposed the catastrophic fragility of Europe’s dependence on imported natural gas. Germany, which had built Nord Stream 2 to increase its reliance on Russian pipeline gas, found itself scrambling to secure liquefied natural gas shipments at astronomical prices while restarting coal plants it had recently pledged to close. The lesson was stark: energy independence is a strategic imperative, not a slogan. Nuclear fuel is energy-dense, easy to stockpile, and sourced from geopolitically stable countries — Australia, Canada, and Kazakhstan supply the bulk of global uranium (World Nuclear Association, 2024). A single fuel load for a reactor lasts eighteen to twenty-four months. A country with a fleet of nuclear plants and a modest uranium reserve can operate for years without imports. No pipeline can be turned off. No tanker can be diverted.

Third, the insatiable appetite of artificial intelligence. The major technology companies did not embrace nuclear out of environmental conviction. They embraced it because they had no alternative. A large language model training run can consume tens of megawatts for weeks. A hyperscale data centre campus can draw 500 megawatts to a gigawatt — the output of a large nuclear plant — continuously. These facilities cannot tolerate intermittency. When Microsoft’s engineers modelled the power requirements for their next generation of AI infrastructure, the conclusion was inescapable: they needed baseload generation that could deliver power around the clock, regardless of weather, at a single interconnection point. Solar and wind, however cheap per kilowatt-hour in favourable conditions, could not guarantee that (Satya Nadella, interview, 2024).

Fourth, new technology. The nuclear industry is no longer confined to the gigawatt-scale pressurised water reactors of the 1970s. A new generation of small modular reactors — factory-built units producing 50 to 300 megawatts each — promises to reduce construction risk by shifting fabrication from bespoke on-site construction to standardised manufacturing. Rolls-Royce SMR in the United Kingdom, backed by £500 million in government funding, aims to deploy its first unit by the early 2030s. NuScale Power in the United States received the first-ever SMR design certification from the Nuclear Regulatory Commission in 2023 — though its first project, at Idaho National Laboratory, was subsequently cancelled due to cost overruns, a reminder that the technology is promising but not yet proven at commercial scale. GE Hitachi’s BWRX-300 has attracted interest from Canada, Poland, and the Czech Republic. China’s HTR-PM — a high-temperature gas-cooled reactor — became the world’s first operational Generation IV reactor in 2023 (World Nuclear News, 2023).

Fifth, fusion is no longer a punchline. In December 2022, the National Ignition Facility at Lawrence Livermore National Laboratory achieved fusion ignition for the first time — producing more energy from a fusion reaction than the lasers delivered to the target. The achievement was a scientific milestone, not a commercial one: the facility consumed far more wall-plug energy than it produced. But it demonstrated that fusion is physically possible, not merely theoretically elegant. Meanwhile, Commonwealth Fusion Systems, an MIT spin-out backed by $2 billion in private capital, is building SPARC, a compact tokamak that aims to demonstrate net energy gain by 2027. The ITER project in southern France — a 35-nation collaboration that has been under construction since 2010 — continues its glacial progress toward first plasma. Fusion remains decades from commercial deployment. But the psychological shift matters: the energy source that powered the sun was no longer confined to the realm of “always thirty years away.”

These five forces converged at the COP28 climate summit in Dubai in December 2023, where 22 nations — including the United States, France, the United Kingdom, Japan, South Korea, and the United Arab Emirates — signed a declaration pledging to triple global nuclear energy capacity by 2050 (COP28, 2023). It was the first time nuclear had featured prominently in a COP declaration. The atom had returned to the diplomatic mainstream.

Winners and Losers — The New Nuclear Geography

The nations that positioned themselves for the nuclear renaissance — or never left in the first place — are pulling ahead. Those that turned their backs are paying the price.

France is the obvious winner, though “winner” implies it did something new. France simply never stopped. Its 56 reactors generate roughly 65 to 70 per cent of the country’s electricity, giving it the lowest carbon intensity of any major industrial economy in Europe and among the lowest electricity prices for industrial consumers on the continent. When the rest of Europe scrambled for alternatives after Russia’s invasion of Ukraine, France had a fleet of reactors that required no gas, no coal, and no geopolitical anxiety. President Macron announced in 2022 that France would build six new EPR2 reactors and potentially eight more, with the first operational by 2035 (Élysée, 2022). EDF, the state utility, has also positioned itself as the contractor of choice for other European nations — including the Czech Republic, which selected the EPR1200 for its Dukovany expansion.

China is building nuclear capacity at a pace that recalls France in the 1970s — but at a scale that dwarfs it. As of early 2026, China has 57 operational reactors and 28 under construction, with plans for approximately 150 reactors by 2035 (World Nuclear Association, 2025). China is constructing reactors in five to six years, roughly half the time of Western projects, using standardised designs and a streamlined regulatory process. The Hualong One — China’s domestically designed pressurised water reactor — is being exported to Pakistan, Argentina, and potentially Saudi Arabia. China has also connected the world’s first Generation IV gas-cooled reactor to the grid and is pursuing sodium-cooled fast reactors and molten salt designs. While the West debates planning permissions, China pours concrete.

The United Kingdom has re-embraced nuclear after a decade of ambivalence. Hinkley Point C, the first new nuclear plant in a generation, is under construction in Somerset — though its budget has ballooned from £18 billion to over £33 billion and its completion date has slipped from 2025 to 2031 (EDF Energy, 2024). The government approved Sizewell C in Suffolk, using the same EPR design, and established Great British Nuclear in 2023 as a dedicated body to accelerate deployment. The SMR competition, centred on the Rolls-Royce design, is intended to deliver smaller, cheaper units that can be sited on former coal plant locations, reusing existing grid connections. Britain’s ambition is clear. Its execution remains to be tested.

The United States, which still operates the world’s largest nuclear fleet at 93 reactors, is experiencing a quiet revival. Several reactors that had been scheduled for retirement have received licence extensions. The Inflation Reduction Act of 2022 included production tax credits for existing nuclear plants — the first federal nuclear subsidy in decades — acknowledging that keeping reactors running is the cheapest and fastest way to maintain zero-carbon baseload generation. The Vogtle expansion in Georgia, which added two AP1000 reactors at a cost of $35 billion (more than double the original estimate), was a cautionary tale in construction management — but the reactors are now operational and producing 2.2 gigawatts of carbon-free power. The Department of Energy has funded demonstration projects for advanced reactor designs, including TerraPower’s Natrium sodium-cooled reactor in Wyoming, backed by Bill Gates.

Japan is the most poignant case. The country that once operated 54 reactors — meeting roughly 30 per cent of its electricity needs — shut all of them after Fukushima. In the decade that followed, Japan burned vast quantities of imported liquefied natural gas and coal, driving up electricity costs and carbon emissions simultaneously. Slowly, painfully, Japan has been restarting its fleet. As of 2025, twelve reactors have returned to service, with more in the pipeline. Prime Minister Kishida declared in 2022 that Japan would maximise nuclear energy use, including the construction of next-generation reactors — a political reversal that would have been unthinkable five years earlier (Government of Japan, 2022).

And then there is Germany — the cautionary tale that future energy historians will study with the same bewildered fascination that military historians reserve for the Maginot Line. Germany shut its last three nuclear reactors on 15 April 2023, in the midst of an energy crisis triggered by its own dependence on Russian gas. The reactors were functioning perfectly. They were producing roughly 6 per cent of Germany’s electricity at near-zero marginal cost and zero carbon emissions. They were shut down because a political commitment made in 2011, in the emotional aftermath of Fukushima, had acquired the force of religious doctrine. The Green Party, the junior coalition partner, refused to countenance an extension. The reactors were switched off. Coal plants were switched on. Germany’s electricity sector emissions rose. Its electricity prices — already among the highest in Europe — remained elevated. Its industrial competitiveness suffered. BASF, the world’s largest chemical company, announced it would shift investment to China, where energy was cheaper (BASF, 2022). The Energiewende had spent approximately €500 billion over two decades and delivered a grid that was more expensive, more carbon-intensive, and more dependent on weather than France’s nuclear fleet — which had cost a fraction of the sum in inflation-adjusted terms (Fraunhofer ISE, 2023).

The German experience is not merely a policy failure. It is a parable about what happens when ideology displaces engineering. The Greens did not oppose nuclear power because the data supported their position. They opposed it because nuclear had been woven into the founding mythology of the German environmental movement — a mythology rooted in Cold War anxiety, Chernobyl trauma, and a romantic vision of distributed, small-scale renewable energy that was always more aesthetic than practical. When the numbers arrived, the mythology held.

The Atom Does Not Care About Your Feelings

The nuclear renaissance is not a matter of ideology. It is a matter of physics, economics, and arithmetic. The world consumes approximately 180,000 terawatt-hours of primary energy per year, and that figure is rising as developing nations industrialise and artificial intelligence devours electricity at rates that would have seemed fantastical a decade ago (IEA, 2023). Decarbonising even a fraction of that supply without nuclear power is not merely difficult — it is, in the view of virtually every credible energy modelling exercise, either impossible or ruinously expensive.

The nations that understood this — France, which committed in the 1970s; China, which committed in the 2010s; the United States and United Kingdom, which are recommitting now — will meet their climate and energy security objectives. They will have affordable, reliable, low-carbon electricity. Their industries will remain competitive. Their data centres will hum.

The nations that delayed by a decade — Germany being the most prominent, though not the only, example — will pay the price in higher emissions, higher electricity costs, and diminished industrial competitiveness. They will discover, as Germany already has, that the market does not reward good intentions, and that a solar panel in Brandenburg produces rather less electricity in January than the brochure suggested.

The atom is returning whether the Greens like it or not. Microsoft, Google, and Amazon have no interest in ideological purity; they have an interest in kilowatt-hours. China has no interest in Western environmental sensibilities; it has an interest in energy dominance. France has no interest in relitigating the 1970s; it has an interest in selling reactors to every European nation that spent thirty years pretending it did not need them.

The question is not whether nuclear power will play a central role in the global energy system of the mid-twenty-first century. The question is which nations will have the courage — and the institutional competence — to build it in time.