The Nuclear Shortage Is a Balance-Sheet Problem

The United States cannot build the nuclear plants it already knows how to build, so it is signing up for ones nobody has ever built. That inversion — hyperscalers turning to unbuilt designs, apparently because the finished ones proved financially ruinous — is the story.

Hyperscalers are not picking TerraPower's Natrium and other unproven designs because they love risk. They are picking them because the only recently finished large-scale build, Vogtle Units 3 and 4, ran over its original schedule, according to public reporting, and cost roughly $36 billion — more than double the original estimate — and left Georgia Power customers carrying roughly 25 percent higher rates, according to public reporting. The Westinghouse AP1000 sitting on the shelf is finished, licensable, and has been built to completion twice. No one will order it.

The freeze is a balance-sheet problem. When a first-of-a-kind reactor runs over, the overruns land on a utility's income statement, which is also a regulated rate base, which is also a politician's problem. Nuclear finance specialists and utility risk managers have widely noted that there is no commercial cost-overrun insurance product for nuclear, no pooled first-of-a-kind risk vehicle, no standard way to spread the pain across hyperscaler offtakers, and no settled regulated-asset treatment for novel designs. The Vogtle hangover is what taught every utility and every regulator that the next first mover pays the full price.

BloombergNEF counts roughly 51 gigawatts of cooperation announcements between data centers and nuclear companies, per BNEF research. Almost none of it is a binding power-purchase agreement. That gap is the shape of the market: ambition is cheap, capital is gated, and capital is gated because the product class that would insure it does not exist.