The Anti-Nuclear Reflex

Chernobyl, 1986: 28 people died of acute radiation sickness, and 2 more in the explosion itself. Upper-bound projections that fold in long-term cancer over the following decades reach into the low thousands, and even those figures are still debated. Fukushima, 2011: one confirmed radiation death. The tsunami that caused the meltdown killed roughly 19,500 people — the reactor was not the lethal event that day. Three Mile Island, 1979: nobody was hurt.

Set that entire history against fossil fuels, which kill about 8.7 million people every year from fine-particulate air pollution alone (Vohra et al., 2021). Coal and gas take more lives in a single afternoon than nuclear has in seventy years of operation. That is the real energy tragedy, and it does not photograph — it arrives as a slightly worse asthma ward, a few thousand extra heart attacks, a town's cancer rate ticking up. Nothing to march against. Just the background hum of normal.

Run the survey experiment and the gap shows up cleanly: show people the risk statistics for different energy sources with the labels stripped off, and they back far larger nuclear fleets than they say they want once the word "nuclear" is attached. The statistics don't scare people. The word does.

Part of it is how the two death tolls arrive. A meltdown is dramatic, rare, and televised for a week. Particulate deaths are invisible and statistical, smeared across millions of quiet cases that never make a headline. The mind grabs the vivid thing and shrugs off the constant one — exactly backwards from what the data would tell it to do.

The rest is history. One word covers both the bomb and the power plant, so the brain files a boron-moderated steam kettle in the same drawer as Hiroshima. That association is doing most of the emotional work, and it has nothing to do with how a reactor actually runs.

Nuclear's real weakness is money and time, not safety. New plants have been expensive and slow — historically ten to twenty years to build, at a cost per kilowatt-hour well above solar or gas. That part is fair, and it's the strongest argument against betting the grid on brand-new reactors. It's also why the current wave is mostly restarts and uprates rather than greenfield builds: when the concrete is already poured, the economics change.

The waste is real, and it gets talked about as if it were unsolved. In volume it is tiny — all the spent fuel the United States has ever produced would fit on a single football field a few yards deep — and the engineering of storing it has, to date, killed no one. The hard part has been political, not technical. And small modular reactors, for all the newsflow, won't add meaningful power before about 2030 (IEA); treat any earlier contribution as a surprise, not a plan.

None of that makes nuclear a silver bullet. It makes it one of a handful of tools that actually work, carrying a real but narrow set of costs — a very different thing from the civilization-ending menace the reflex imagines.

When the data centers came for the grid, the math stopped being optional. Three Mile Island — yes, that one — is being restarted specifically to feed them. Vistra signed twenty-year deals with Meta for more than 2,600 megawatts of existing nuclear. Amazon wired a data center straight to the Talen reactor next door. These are seventeen-to-twenty-year off-take agreements, the kind that only get signed when both sides believe the plant will still be running in 2045.

The policy reversals followed the same logic. Since 2023, Belgium, Italy, Japan, Sweden, Denmark, Poland and Romania have all rolled back anti-nuclear positions under energy pressure. Not on principle, and not because a documentary changed their minds — on arithmetic. The demand curve made the reflex too expensive to keep.