Nuclear-powered AI isn’t going to happen overnight

Yes, $80 billion will help fund the new reactors, but building takes at least 10 years.

- By WILL WaDE

NATHAN HOWARD Bloomberg A POWER substation near Virginia's Data Center Alley. The AI boom has supercharged electricity use.

After a generation of stagnation, interest in U.S. nuclear energy is sky high, thanks to artificial intelligence’s voracious energy needs, Nuclear companies are stock market darlings, and the Trump administration recently unveiled an $80-billion-plus commitment to help fund new reactors.

There’s just one problem: Hardly anything will be ready to plug in for a decade.

In theory, $80 billion can buy enough reactors to power all of Virginia’s Data Center Alley. In reality, traditional reactors take 10 years or more to build, while a highly anticipated wave of small, modular reactors has yet to produce commercial power. That means the only plants that will go into service anytime soon are a handful of shuttered facilities that can be restarted.

“There’s a lot going on, and nothing is going on,” said BloombergNEF's head nuclear analyst Chris Gadomski.

That dissonance underscores how radically, and rapidly, the U.S. energy system has transformed in recent years. Not long ago, utilities were shutting down costly nuclear plants and replacing them with cheaper natural gas-fired and renewable plants. The AI boom has since supercharged electricity consumption, straining power supplies and driving up prices. But while nuclear is suddenly in demand again, logistics and technology constraints threaten to thwart development even with President Trump’s moves to streamline permitting and commit billions of dollars in funding.