“This project will create jobs, spur local innovation, and advance American leadership in energy technology,” Urvi Parekh, head of global energy at Meta, said in a statement. “By investing in baseload nuclear energy, we’re helping build a resilient and sustainable future for our communities.”

It’s not unusual for utilities to negotiate long-term contracts for fuel for reactors. But this is the first known incident where a hyperscaler is purchasing the fuel that will generate the electrons it plans to buy, says Koroush Shirvan, a researcher at the Massachusetts Institute of Technology.

“The Oklo model that they have advertised is that they build, own, and operate,” says Shirvan. “But I’m trying to think of any other customers who provide fuel other than the U.S. government. I can’t think of any.”

Oklo emerged in the past year as the poster child for a possible revolution in the US on how nuclear plants are built. Until recently, the US hadn’t started and completed any new reactors in a generation. By the time the only new machines came online at a Southern Company power plant in northern Georgia in 2023 and 2024—a pair of 1,100-megawatt Westinghouse AP1000s, the leading design for a traditional reactor in the US—the project was billions of dollars over budget and more than half a decade late. But the second unit came in roughly 30 percent cheaper than the first, a sign of the efficiencies gained by repeating the same design.

To fix this problem, a growing faction in the nuclear industry proposed shrinking the size of reactors, so that building a 1,000-megawatt plant would require constructing multiple reactors of the same size, ultimately bringing down the cost. Many of those companies, including NuScale Power and GE Vernova-Hitachi Nuclear Energy, focused on building shrunken-down versions of the water-cooled reactors that make up all of America’s fleet of 94 units. But Oklo and rivals such as X-energy, Google-backed Kairos Power, and Aalo Atomics instead looked for a totally clean slate, seeking to commercialize experimental reactor models that use coolants such as sodium, molten salt, or high-temperature gas rather than water.

That kind of design required a different kind of fuel like HALEU, one that could burn up more of the energy locked in the uranium than traditional reactors could. The trouble was that the on...