Nuclear
The safest way we have ever generated electricity, by a very wide margin, and one of the most expensive. Both of these are true and almost nobody will say both.
The effect compounds within years. Put it in place and it keeps working.
Origins
Nuclear power was born in a bomb, and it has never entirely escaped it.
The first controlled chain reaction was achieved under a squash court in Chicago in 1942, as part of the Manhattan Project. The technology arrived in the world as a weapon, and the civilian programme that followed — Eisenhower’s “Atoms for Peace” in 1953 — was in part an attempt to give the atom a second reputation.
For about twenty years it worked. The 1960s and 70s were an era of genuine nuclear optimism: electricity “too cheap to meter,” France building an entire national grid on it in fifteen years, and doing it so well that French electricity is still among the cleanest in the developed world.
Then came Three Mile Island in 1979, and Chernobyl in 1986, and Fukushima in 2011. The public verdict was rendered, and it was rendered on television.
What makes nuclear’s history so strange is the gap between that verdict and the numbers. Chernobyl was a genuine catastrophe, caused by a reactor design nobody would build now, operated recklessly, in a state that lied about it. Fukushima killed almost nobody through radiation — the deaths came from the tsunami and from the panic of the evacuation itself. And meanwhile, invisibly and continuously, coal smoke has been killing people at a rate that would constitute a Chernobyl every few weeks, and it does so without a single news bulletin.
Nuclear is the technology we are most afraid of and the one that has hurt us least. That is a fact about human psychology, not about physics.
What it actually is
A nuclear plant splits uranium atoms, releases heat, boils water, and drives a steam turbine. It is, underneath, a very sophisticated kettle. It emits no carbon dioxide while running, produces enormous amounts of power from a very small quantity of fuel, and runs continuously at capacity factors above 90% — the highest of any source, in any weather, day or night.
It is firm power: the thing renewables cannot yet fully provide and the thing grid operators most want.
The problems are not the ones most people think.
Safety is not the problem. The mortality figures are stark and consistent: roughly 0.07 deaths per terawatt-hour for nuclear, against 24.6 to 100 for coal. Nuclear is in the same safety band as wind and solar, and hundreds of times safer than the fuel it would displace. This includes Chernobyl. This is not a fringe claim; it is the mainstream estimate.
Waste is not really the problem either. It is a solvable engineering question that has been made politically insoluble. The total volume is small, it is contained, and Finland has actually built a permanent geological repository while everyone else argued.
Cost and time are the problem. Western nuclear construction has been a financial disaster for two decades: chronic overruns, decade-long build times, and projects that bankrupt their builders. That is the honest case against it, and it is a strong one.
The numbers
Safety, which is not close. Nuclear causes roughly 0.07 deaths per terawatt-hour generated. Coal causes between 24.6 and 100, depending on the estimate. That is a factor of several hundred to over a thousand, and it includes Chernobyl. Nuclear sits alongside wind and solar as one of the safest ways humans have ever made electricity.
Capacity factor. Around 93% for nuclear, against roughly 24% for solar and 35–50% for wind. A nuclear megawatt delivers several times more electricity per year than a solar megawatt, which is why simple cost-per-watt comparisons mislead.
Cost, which is brutal. Nuclear LCOE, including SMRs, has been estimated at US$141–221/MWh, against US$24–96/MWh for utility solar and wind. Nuclear construction overruns average 117%, against 71% for hydro, 8% for wind and 1% for solar. That last comparison is the whole argument in one line.
The SMR promise, and the SMR reality. Small modular reactors are meant to escape this by being factory-built rather than bespoke. First-of-a-kind SMRs currently cost around US$8,500–10,500/kW — roughly eight times utility solar per megawatt — with LCOE around US$80–150/MWh. The flagship American project, NuScale’s Carbon Free Power Project, saw costs rise from $5.3 billion to $9.2 billion and was cancelled. The theory requires reaching mass production. Nobody has yet.
The system argument. Modelling by Sepulveda and colleagues found that including firm low-carbon sources in a fully decarbonised grid lowered total system costs by 10–62%. Nuclear may be expensive per megawatt-hour and still make the whole system cheaper. This is the strongest argument for it and it is rarely made.
Where it stands. Nuclear supplies about 10% of global electricity and generation is at historic highs. But the IEA expects solar PV alone to overtake nuclear in 2026.
Why it matters
We are going to try to do something unusual here, which is to be genuinely undecided in public.
Nuclear power is the most emotionally charged subject in energy, and both tribes have made themselves ridiculous. The anti-nuclear movement has spent fifty years fighting the safest source of electricity we have ever built, and in doing so has helped keep coal plants running that kill people continuously and invisibly. Germany closed its reactors and burned lignite instead, and people died of the air, and the emissions went up. That is not a hypothetical; it happened, and the movement that caused it has never really reckoned with it.
And the pro-nuclear side has spent the same fifty years insisting that the next reactor will be on time and on budget, and it never is, and the costs are not a conspiracy by environmentalists — they are real, and they have bankrupted companies, and pretending otherwise is not advocacy but wishful thinking.
Here is what we think, and you are free to disagree.
Existing reactors should keep running. They are built, they are paid for, they are safe, they are clean, and closing them means burning gas. Closing a working reactor to build a solar farm is not progress; it is standing still while spending money.
New nuclear has to earn it. Not through fear of it, and not through faith in it. Through cost, on time, demonstrated. If SMRs can do that, they will be part of the answer. If they cannot, the money is better spent on wind, solar, storage and grid.
That is an uncomfortable position that satisfies nobody, and we think it is the honest one.
What it actually takes
Building the same thing twice. The West’s nuclear failure is fundamentally a failure to standardise. France built one design, over and over, and it worked. America and Britain built bespoke plants, each one a prototype, each one relearning the lessons of the last. Nuclear is only cheap if you build a fleet. It is ruinous if you build a monument.
Regulatory reform that does not reduce safety. This is genuinely difficult and it is where the debate is most dishonest. Licensing takes years and costs billions, and much of it is genuinely necessary. Some of it is not. Distinguishing the two requires more rigour than either side has shown.
Financing, which is the real killer. A plant that takes ten years to build accumulates interest for a decade before earning a penny. At an 8% cost of capital, a seven-year build adds roughly 50% to the overnight cost. Nuclear’s problem is arithmetic, and the arithmetic is about time.
Actually solving the waste politics. Finland has built a permanent geological repository. The technical problem is solved. Everywhere else it remains stranded in political limbo, and every year of delay is used as an argument against the technology it is a delay in.
And letting go of the fear. Not the caution — the caution is right. The fear. The numbers on safety have been consistent for decades and almost nobody believes them, and that gap is doing measurable harm.
Where it matters most
France is the standing proof that it can be done: a national grid decarbonised in about fifteen years, still among the cleanest electricity in the developed world, built by standardising on a small number of designs.
Germany is the standing proof of the opposite. Reactors were closed on schedule after Fukushima, lignite was burned in their place, emissions rose and air quality worsened. It is the single most instructive policy error in European energy and it was made by people who thought they were being green.
China and South Korea are the only places currently building nuclear at reasonable cost and on schedule, and they are doing it by building the same design repeatedly. That is not a coincidence.
The United States has the largest fleet in the world, mostly built in the 1970s, mostly running well, and mostly approaching end of licence. The most valuable nuclear decision available to America is not building new reactors. It is not closing the ones it has.
Finland has done the thing everyone said was impossible and built a permanent waste repository. It is worth knowing that this happened, because the debate proceeds as though it did not.
How to tell it’s being done well
Is it a repeat build or a prototype? This single question predicts the cost outcome better than anything else. Fleets are affordable. One-offs are catastrophic.
Are the costs honest? Nuclear projects have a long and consistent history of optimistic initial estimates. An average overrun of 117% is not bad luck; it is a pattern, and any new project should be assessed against it.
Is the comparison fair? Comparing nuclear per megawatt to solar per megawatt ignores that nuclear runs at 93% capacity factor and solar at 24%, and that firm power lowers whole-system cost. Comparing them per megawatt-hour without accounting for storage ignores the reverse. Both sides do this, constantly.
Is the alternative honest? Closing a working reactor means something replaces it. If that something is gas, the emissions are worse and the deaths are higher. Ask what the counterfactual actually is.
What you can do
Anyone
- Look up the deaths-per-terawatt-hour figures for yourself. Nuclear is around 0.07; coal is between 24.6 and 100. That comparison includes Chernobyl and it surprises almost everybody.
- Distinguish between keeping existing reactors open and building new ones. They are completely different questions with different economics, and conflating them is the most common error in this debate.
Policymakers
- Do not close working reactors. They are built, paid for, safe and clean, and what replaces them is usually gas. Germany is the cautionary tale.
- If you build, build a fleet of one design. Bespoke nuclear is ruinous; standardised nuclear is affordable, and France proved it.
- Solve the waste politics. Finland built a permanent repository. The technical problem is solved and the political one is being used as an argument against the technology it is delaying.
Investors
- Nuclear's core problem is financing, not physics. A ten-year build accumulates interest for a decade before earning anything, and at 8% cost of capital a seven-year build adds roughly 50% to overnight cost.
- SMR economics depend entirely on reaching mass production. First-of-a-kind costs are around eight times utility solar per megawatt. Nobody has crossed that curve yet.
Environmental advocates
- The anti-nuclear movement helped keep coal plants running for fifty years. That is not an accusation; it is an outcome, and it deserves an honest reckoning rather than a defensive one.
- Opposing new nuclear on cost is defensible. Opposing it on safety is not supported by the evidence, and making that argument damages your credibility on everything else.
Who is working on this
We are researching which organizations in our directory of 8,493 actively work on this solution, and we only list an organization once we have verified it. That research is ongoing. In the meantime, search the directory yourself:
Questions
Is nuclear power safe?
It is among the safest ways humans have ever generated electricity, and the figures are not close. Nuclear causes roughly 0.07 deaths per terawatt-hour, against 24.6 to 100 for coal. That is a factor of several hundred to over a thousand, and it includes Chernobyl. Nuclear sits alongside wind and solar in the safest band, and the gap between that fact and public perception is one of the widest in any policy area.
What about Chernobyl and Fukushima?
Chernobyl was a genuine catastrophe, caused by a reactor design nobody would build today, operated recklessly, in a state that concealed it. Fukushima killed almost nobody through radiation; the deaths came from the tsunami and from the evacuation itself. Both were serious. Neither changes the mortality comparison with coal, which kills continuously and invisibly and never makes the news.
So why is nuclear controversial?
Cost and time, if the argument is being conducted honestly. Western nuclear construction has been a financial disaster: overruns averaging 117% against 8% for wind and 1% for solar, decade-long build times, and projects that have bankrupted their builders. That is a strong argument and it has nothing to do with safety.
What about nuclear waste?
It is a solved engineering problem that has been made politically insoluble. The total volume is small and it is contained. Finland has actually built a permanent geological repository while everyone else was arguing. The technical answer exists; the political will mostly does not.
Will small modular reactors fix the cost problem?
They might, and they have not yet. SMRs are meant to be factory-built rather than bespoke, escaping the cost disease of megaprojects. But first-of-a-kind SMRs currently cost around US$8,500 to 10,500 per kilowatt, roughly eight times utility solar per megawatt, and the flagship American project saw costs rise from $5.3 billion to $9.2 billion before being cancelled. The theory requires reaching mass production, and nobody has crossed that curve.
Should existing reactors be kept running?
On the evidence, yes. They are built, paid for, safe and clean, and closing them means something else generates that electricity, which in practice is usually gas. Germany closed its reactors after Fukushima, burned lignite instead, and emissions rose. Closing a working reactor to build a solar farm is not progress; it is standing still while spending money.
Sources
- Project Drawdown - Nuclear Power (Drawdown Explorer) Framework and classification. Cited, not reproduced.
- Our World in Data - What are the safest and cleanest sources of energy?
- Sepulveda et al. (2018), Joule - The role of firm low-carbon electricity resources in deep decarbonization
- IEA - Electricity 2026
- IEEFA - analysis of NuScale Carbon Free Power Project cost escalation
- Eash-Gates et al. (2020), Joule - Sources of cost overrun in nuclear power plant construction
The solution taxonomy follows the framework popularised by Project Drawdown. The analysis above is our own; for their carbon modeling and rankings, visit them directly.