The US bets $17.5 billion on nuclear power — a decade before it can deliver

Climate & Energy 3 min 64 sources

Washington poured conditional loans into ten new reactors to feed the AI boom, even as it killed a nearly-built wind farm and battery storage set records. The bet is on the slowest power source to answer the fastest-moving demand.

Key takeaways

The US committed $17.5 billion in loans for ten new nuclear reactors to feed soaring AI data-center demand — but nuclear plants take a decade-plus to build, and America's last ones ran seven years late and billions over budget.
The same week, the government moved to kill a nearly-ready offshore wind project while battery storage hit record installs and drew $5 billion in private money — fast power sources that arrive in months, not years.
The bet is structural: matching the fastest-moving demand in energy with the slowest, most committing supply, and locking in that choice before anyone knows what the 2036 grid will actually need.

The U.S. Department of Energy announced $17.5 billion in conditional loans on Tuesday to help utilities buy the parts for ten new large nuclear reactors [3][5]. The money pays for the long-lead components — reactor vessels, steam generators — that “can take years to secure” [5]. Energy Secretary Chris Wright said it could pull the goal of ten reactors under construction by 2030 forward by three years [5]. The driver, in his own words: data center “hyperscalers” — the tech giants running the cloud and AI computing whose electricity demand is “soaring” [5].

Why nuclear, and why now

Nuclear power gives you a lot of always-on electricity with almost no carbon. That is the appeal: AI data centers run flat out, day and night, and want power that never blinks. Nuclear delivers exactly that profile.

The catch is time. The last reactors America finished, at Georgia’s Vogtle plant, came in seven years late and billions of dollars over budget [5]. The new plan uses Westinghouse AP1000 reactors, with the government arranging financing and permits in exchange for a 20% share of future profits [5]. Each project needs a utility partner to commit $500 million before the federal money unlocks [5]. Trump’s stated aim is to quadruple U.S. nuclear capacity to 400 gigawatts by 2050 [5] — a quarter-century horizon for a demand surge happening this year.

The thing it killed, and the things it didn’t fund

The same week, the picture on faster power sources looked very different.

California said it intends to sue the Trump administration over a deal to end a major offshore wind project — billions of dollars of capacity, much closer to delivering than any new reactor [1]. Offshore wind takes years too, but this one was already in motion.

Meanwhile, the market moved on its own. U.S. energy storage hit a record in the first quarter of 2026 [2]. Independent firm NatPower and Tesla signed a deal for the first phase of a $5 billion battery buildout — 25 gigawatt-hours across Italy and Britain, part of a 100 GWh plan [10]. Batteries don’t make electricity; they store it and release it when demand spikes, smoothing out wind and solar. They take months to install, not a decade.

So three power strategies sat side by side this week: a decade-long nuclear bet getting federal billions, a near-ready wind project getting cancelled, and fast-moving storage getting private money and no headlines.

The bet underneath the bet

A government choosing nuclear isn’t just choosing clean power. It’s choosing a long waiting period, a fixed design, and a commitment locked in before anyone knows what 2036’s grid actually needs. If AI demand keeps climbing the way Wright expects, the reactors arrive into a hungry market. If demand cools, or storage and cheaper renewables fill the gap first, the country has spent billions on capacity that took ten years to show up.

Wright called it “not a risky endeavor” [5]. The history of building these plants — late, over budget, and capital-intensive enough that private money has avoided them for decades [5] — says the risk is mostly about time.

For an ordinary person, the stakes land on the electricity bill. Nuclear’s costs, when they overrun, get spread across ratepayers and shareholders [5]. The promise is stable power and stable prices a decade out. The gamble is paying for it now, on a guess about a demand curve that’s moving faster than concrete can be poured.

02 · Lesson · why it matters

When you answer a fast problem with a slow solution

The hardest choices aren't between good and bad options — they're between things that move at different speeds, where the one you commit to today won't arrive until the question has changed.

Two clocks, not one

The demand is moving on a fast clock. AI data centers go from plan to running in a year or two. The companies building them want power now, and more of it every quarter.

The answer being funded runs on a slow clock. A nuclear reactor is a ten-year project at best, and America’s recent record is worse — seven years late, billions over budget. The $17.5 billion announced this week buys parts that “take years to secure,” for plants that won’t feed a single server before the 2030s.

When a fast problem and a slow solution sit in the same decision, the speeds matter more than the merits. Nuclear power is genuinely good at what it does. That isn’t the question. The question is whether a ten-year build is the right tool for a demand curve nobody can read past next year.

The gap is where the risk lives

Between the moment you commit and the moment it arrives, the world keeps moving — and you can’t take the commitment back.

Pour the foundation, order the reactor vessel, sign the 20% profit deal, and you’ve fixed your answer in concrete. If the AI surge keeps climbing, the timing works and the bet pays. If demand cools, or if batteries and cheap renewables quietly fill the gap first — storage just hit a record, and a $5 billion battery deal landed the same week — then the slow thing arrives into a world that already solved the problem another way.

The longer the gap, the more the bet is really a bet about the future, not the present. You aren’t choosing the best power source. You’re choosing the demand curve you think 2036 will have.

Why we reach for the slow thing anyway

It feels like the serious choice. Big, permanent, decisive — a reactor is something you can point to. The fast options look small and hedged by comparison: a battery farm, a wind project, a contract you can adjust. Decisiveness reads as strength, and a ten-year commitment looks bolder than a flexible one.

But under uncertainty, the flexible option often is the stronger one — precisely because it doesn’t gamble on a guess. Killing the nearly-ready wind project to fund the decade-long reactor isn’t choosing the better power. It’s choosing the more committing one, and calling the commitment confidence.

Who else is standing in the gap

The people who live with this bet aren’t in the room where it’s made.

A ratepayer a decade from now inherits the cost overruns if the plants run late — that’s how Vogtle’s billions got paid. A grid operator in 2034 inherits a fixed slab of capacity that can’t be dialed down if it turns out the country needed something nimbler. The household paying an electricity bill is downstream of a guess made years earlier about a demand curve that was still being drawn.

None of them chose the speed mismatch. They live inside the wait.

What the whole looks like

Most big decisions are not really about which option is best. They’re about which clock you’re on, and whether the thing you’re committing to will still fit the problem by the time it shows up.

That is the humbling part. The people making this bet are not foolish — nuclear is real, clean, and powerful. But no one in the room can see what 2036 needs, and the decision locks in anyway. The honest position isn’t certainty in either direction. It’s holding the commitment a little more loosely than a ten-year contract allows — knowing that the gap between decide and arrive is exactly where the future hides.

03 · Lab · your turn

The Power You Pick Today

Commit to a power source under hidden demand, then watch the gap between deciding and arriving decide the outcome.

04 · Hope · carry this

No single bet has to carry the whole future. The same week one slow plan got its billions, faster ones quietly broke records — and a world with that many paths to clean power is a world still finding its way forward.