For decades, nuclear power was the ghost at the feast of clean energy. Expensive, slow, haunted by Three Mile Island and Fukushima. Environmentalists wrote it off. Investors ran from it. Regulators buried it in red tape. And then something strange happened.
Four reactors just hit criticality in the United States. Not old ones being relicensed—new builds, actually coming online. And the reason they exist has almost nothing to do with climate marches or carbon targets. It has everything to do with the cold, insatiable hunger of a datacenter.
Nuclear energy isn’t being revived by environmentalists. It’s being rescued by the cold, hard math of AI’s electricity bill.
You’ve probably noticed the headlines: Google, Microsoft, Amazon are signing power-purchase agreements with nuclear startups faster than you can say “small modular reactor.” OpenAI’s Sam Altman is personally funding a nuclear fusion company. The reason is brutally simple. AI training runs consume as much electricity as a small city. And the only source that can run 24/7 without emitting carbon is the one we spent thirty years pretending was dead.
Let me be clear: this is not a compromise. This is a rescue mission. The same tech giants who once promised to run their clouds on 100% solar and wind have hit the wall—literally, because the sun doesn’t shine at night. They need baseload power. They need density. They need a plant that fits on a few hundred acres, not a thousand square miles of solar panels. And they are willing to pay a premium to get it built yesterday.
The irony is delicious: the industry that killed nuclear for a generation—cost overruns, regulation, NIMBY lawsuits—is now being crushed by the sheer weight of demand from the most hyper-capitalist sector on Earth.
I saw this firsthand in a briefing last month: a utility executive told me their biggest headache isn’t anti-nuclear activists anymore. It’s a tech company that wants to co-locate a datacenter inside the plant’s fence. “They don’t care about the waste,” he said. “They care about latency and uptime.” That’s the new reality. The enemy of nuclear wasn’t the greens—it was the banks. And now the banks see a customer with a checkbook that never bounces.
Of course, this isn’t a magic wand. The reactors hitting criticality now are the result of decisions made a decade ago. The regulatory path is still a minefield. But when your buyer is a trillion-dollar company that needs 500 megawatts to train its next model, suddenly the risk of building a reactor looks a lot smaller than the risk of not having enough power.
Climate change didn’t save nuclear. Capitalism did. And if you’re worried about the climate, you should be thrilled.
Because here’s the twist: those reactors aren’t just feeding AI. They are baseload machines that run 90% of the time. Once they’re built, they don’t care whether they’re powering a ChatGPT query or a hospital. They just pump electrons into the grid. The AI boom is building infrastructure that will serve everyone—long after the hype cycle ends.
So the next time someone tells you nuclear is dead, ask them who’s paying for the new ones. The answer isn’t a climate activist. It’s a server rack in Virginia. And that server rack is going to save us all.
FAQ
Q: Aren't nuclear reactors still too expensive and slow to help with AI's immediate energy needs?
A: Yes, today's large reactors are expensive and take a decade to build. But tech giants are investing heavily in small modular reactors (SMRs) and advanced designs that promise faster, cheaper deployment. The key is that AI's demand creates a guaranteed revenue stream, which de-risks investment and accelerates regulatory approval.
Q: What's the practical implication for someone who isn't in tech or energy?
A: If AI-driven nuclear succeeds, you get cheaper, cleaner, and more reliable electricity. That means lower bills, fewer blackouts during heatwaves, and a real chance at decarbonizing the grid without relying on intermittent renewables. Your next smartphone or electric car could be powered by a reactor that was justified by a chatbot.
Q: Isn't this just another way for Big Tech to greenwash while consuming ever more energy?
A: The greenwashing critique misses the point. Yes, AI is increasing energy demand—but that demand is going to be met one way or another. If it's met by natural gas, emissions rise. If it's met by nuclear, emissions fall. The fact that profit-driven companies are building nuclear plants is not a moral failing; it's a market signal that clean baseload power finally has an economic champion.