Consider this paradox, and sit with it: the most reliable electricity on the grid is generated by machines that need a constant supply of cold water, and the one thing the world has started to run short of is cold water. In early August, a nuclear plant in Hungary dropped its output to just over 10 percent of capacity and came close to a full shutdown. The plant normally provides roughly 40 percent of the country’s electricity. The cause was not a breakdown, not a strike, not a fuel problem. The river feeding its cooling system had gotten too warm and too low. Regional spot prices responded in the only way markets know how: they jumped to around 500 euros per megawatt-hour.
Say it plain, and it sounds like a joke: the nuclear plant was defeated by a warm summer. And that is the point. The joke is the diagnosis.
We have spent decades arguing about nuclear power — about cost, about waste, about safety — and the argument has quietly assumed one thing nobody mentions: that the plant will have water when it needs it. France learned the same lesson this summer, on a grander scale. Through July, as river temperatures and flows turned against the reactors, the French grid operator curtailed or shut down nuclear capacity at a record pace, even while insisting that supply remained sufficient. By late August, French nuclear output had fallen to about 34 gigawatts, with a wave of strike-related shutdowns and scheduled outages layered on top. The country that built its entire electricity identity on nuclear baseload spent the summer discovering that baseload is only as solid as its coolant.
Let me be honest about my own blind spot here, because I held the standard view for a long time. I treated heat waves as weather stories — human-interest items about beaches and heat advisories. Nuclear power, in my mental ledger, was the boring, dependable anchor of the energy system, the thing that hums along while renewables come and go. The events of this European summer corrected that read with a wet slap. The dependence works both ways: renewables depend on the wind and the sun, and the always-available machine depends on a river that is exactly as reliable as the rain that fell last winter. The reliable thing was never the machine. The reliable thing was the assumption that water would always be there.
That realization took me somewhere I did not expect, and I want to walk you through the reasoning rather than just present the conclusion. I started writing this piece from the price angle — 500 euros per megawatt-hour is a dramatic number, and price spikes are easy to feel. But the deeper I went, the more I realized the price was a symptom, not the story. The story is the cooling loop. Every thermal plant — nuclear, coal, gas — is, at bottom, a machine for boiling water and then condensing it back. The condensing step requires a heat sink, and the heat sink is a river, a lake, or the sea. Take the cold away, and you do not have a plant; you have a very expensive kettle. The price spike was the market’s way of saying what the physics had already said.
Here is the part that made me pause and reconsider what we mean by grid stability. For years, the standard critique of renewables has been that they are intermittent — the sun sets, the wind drops, and suddenly you need something dependable underneath. The European summer of 2026 is the mirror image: the dependable technology turned out to be climate-dependent too, just on a slower and less visible clock. The sun and the wind fail you in hours. The cooling water fails you in seasons, and the failure arrives as a slow squeeze — warmer river, lower flow, reduced discharge limits to protect fish and drinking water — until one day the operator has to throttle the reactor because the water can no longer cool it. Intermittency was never the unique weakness of renewables. It was the universal weakness of anything that needs an environment.
The Lancet puts a human number on what we are discussing in engineering terms: a study covering 2000 to 2019 estimated that around 480,000 people globally die from high heat every year. Let me sit with that number for a moment, because it is easy to read it and move on. That is roughly the population of a mid-sized city, every year, dying of the thing that also slows your power plant. The energy system and the human body are stressed by the same weather, and we tend to discuss them in separate rooms — one in engineering conferences, one in public health warnings. The summer in Europe is a reminder that they are one problem wearing two hats. The heat that makes your river too warm to cool a reactor is the same heat that fills emergency rooms.
Now, the counterargument, and I want to give it a fair hearing because it is not stupid. A defender of the current system would say: this was a bad summer, but grids did not collapse, France insisted supply was adequate, and engineering will fix the cooling problem with dry cooling towers and better discharge management. There is truth in that. The French system did hold, and dry cooling exists, and plants can be hardened. But here is the difficulty with that defense, and it is the difficulty at the heart of the whole climate adaptation argument: the fix is slow, and the summers are arriving faster than the fixes. A dry cooling retrofit on a major plant is a multi-year, multi-billion-euro project. Heat waves are not waiting for the retrofit to finish. The sensible version of the defense is not ‘we are fine’ — it is ‘we are fine, provided we start the retrofits now and the next bad summer arrives late.’ That is a bet, and the climate is declining to take the other side of it.
And here is the uncomfortable part that the price signal hides, because prices do not feel pain — people do. A 500-euro megawatt-hour is an abstraction to the grid operator and a catastrophe to the household that pays a variable tariff. When heat forces a nuclear fleet down, the cheapest fuel disappears from the mix, and the marginal price is set by whatever is left — usually gas, usually expensive. The people who absorb that cost are the ones with the least flexibility: renters, small businesses, pensioners in apartments without cooling. The paradox sharpens. The very technology that was supposed to be the stable, affordable bedrock becomes, on the hottest days, the source of the most volatile prices.
This is where the practical advice has to enter, because a tea-house argument that stops at the paradox is only half a service. The specialists quoted in the reporting on this summer’s events are converging on the same list, and it reads like the maintenance schedule of a building that finally got inspected. First, storage and demand response are not luxuries anymore; they are the shock absorbers that let a grid survive a week of squeezed nuclear output. Second, cooling systems need their own climate upgrade — dry cooling, better intake management, discharge rules that account for hotter baselines rather than assuming yesterday’s river. Third, the grid plan itself has to stop being written around a 20th-century weather assumption. Plan for the hotter summer as the new normal, not as the once-in-a-century anomaly, and the surprises shrink.
Let me reach for an older comparison, because this is not the first time a system discovered its hidden dependence on the ordinary. A century ago, the railways discovered that their schedules were hostage to the snow; the telegraph discovered it was hostage to the wind. Every infrastructure we built assumed a weather pattern that seemed permanent, and every one of them was corrected the hard way. The power grid is simply the latest in that line, and its correction is arriving in the form of a 500-euro price signal and a nuclear plant throttling itself down because its river ran warm. The pattern is ancient; the machine is new.
Say it plain, one more time: the machine we built to be independent of the weather is, at its most critical joint, a weather machine. And that’s the point that the whole debate keeps circling without quite landing on. The argument about whether nuclear power is clean or dirty, cheap or expensive, has been conducted as if the answer were a property of the technology. It is not. It is a property of the environment the technology runs in. Put the same reactor next to a cold, high-flow river, and it hums. Put it upstream of a summer that no longer produces cold water, and it throttles. The technology did not change. The environment did — and the design documents have not caught up. And that’s the point.
The planning implication is uncomfortable but clear. Every grid expansion study, every capacity auction, every demand forecast, will now need a climate column alongside the load column: what does this system look like in the July of a 40-degree summer, with river flows at a fraction of the historical average and cooling limits binding? The engineers I have read on this are converging on the same answer — storage and demand response are the shock absorbers, and cooling infrastructure needs the same climate retrofit that buildings do. The good news is that the tools exist. The bad news is that they are being ordered after the first failures, not before them.
And here is where I land, and I admit I did not expect to land here. The nuclear debate has been conducted for forty years as a binary — nuclear, yes or no. This summer suggests the more honest question is: nuclear, with what cooling assumptions? A plant that is planned around cooling water it may not have in thirty years is not the same asset as a plant planned around a climate-adjusted water budget. The technology is not the whole story; the environment it runs in is part of the design, and the environment has changed while we were arguing.
Consider this the next time you hear someone call nuclear power the reliable backbone of a low-carbon grid: reliable against what? Against a windy week, yes. Against a hot, dry summer, the record now says otherwise. The paradox is that the machine built to free us from the weather remains chained to the weather at its most ordinary point — the water intake. And that is the point, the whole point: we built a civilization-scale machine and handed it a vulnerability so mundane that we forgot to check it. The joke is the diagnosis, and the diagnosis is that our most dependable technology has the same weakness as our most fragile one — it needs the world to behave as it used to.