🔻Clean Energy

Analysis: Big Tech Bets on Nuclear Fission as AI Pushes U.S. Power Demand Higher

Major technology companies are investing in next-generation reactors to secure reliable, around-the-clock energy for their data centers as artificial intelligence accelerates electricity needs.

Analysis: Big Tech Bets on Nuclear Fission as AI Pushes U.S. Power Demand Higher

Representational Photo

BY Donna Joseph

NEW YORK, May 27, 2025 — Artificial intelligence has driven U.S. electricity demand sharply upward after years of stagnation, pushing major technology companies to secure new power sources for their energy-hungry data centers. Increasingly, they are turning to nuclear fission — a technology undergoing a revival after decades of decline.

Unlike nuclear fusion, still in its experimental phase, fission has powered commercial nuclear plants for decades. Companies like Amazon, Google, Meta, and Microsoft are drawn not only to the stable, 24/7 energy output but also to emerging designs that address the drawbacks of older nuclear plants.

Small modular reactors, or SMRs, are at the forefront of this shift. Unlike traditional reactors that generate over 1 gigawatt, SMRs are built in multiple small units, promising flexibility and cost reductions through mass manufacturing — though no such reactor has yet been built in the United States. Even so, major technology companies are investing heavily and signing long-term purchase agreements with a new generation of nuclear startups, several of which are outlined below.

Kairos Power

Google has pledged to purchase about 500 megawatts from Kairos Power by 2035, with the first reactor expected online by 2030. Led by CEO Mike Laufer, Kairos’ reactors use molten fluoride salt for cooling, operating at low pressure for enhanced safety. The Alameda-based company received $629 million in U.S. government funding, including $303 million from the Department of Energy. In November 2024, Kairos secured federal approval to build two 35-megawatt test reactors in Tennessee, with future commercial units expected to reach 75 megawatts.

Oklo

Backed by OpenAI CEO Sam Altman, Oklo focuses on liquid metal-cooled reactors designed to reduce nuclear waste. Altman, who chaired the company until April, took Oklo public in 2023 through a reverse merger with his special purpose acquisition vehicle, AltC. Though Oklo’s initial license application was denied in 2022, the company plans to resubmit in 2025. Meanwhile, Oklo has secured a deal to supply 12 gigawatts to data center operator Switch by 2044.

Saltfoss

Saltfoss, formerly Seaborg, takes a maritime approach, designing Power Barges equipped with two to eight molten salt-cooled reactors. The company has raised nearly $60 million, including backing from Bill Gates, Peter Thiel, and Unity co-founder David Helgason, according to PitchBook. Saltfoss has partnered with Samsung Heavy Industries to build both the ships and reactors.

TerraPower

Bill Gates’ TerraPower is advancing the Natrium reactor, a larger design cooled by liquid sodium and featuring molten salt energy storage. The Wyoming project, which broke ground in June 2024, aims to generate 345 megawatts — smaller than today’s new nuclear plants but larger than most SMR designs. The molten salt system stores excess heat, allowing the reactor to run steadily even when electricity demand is low.

X-Energy

Amazon’s Climate Pledge Fund led X-Energy’s $700 million Series C-1 round last year. The company is developing high-temperature, gas-cooled Xe-100 reactors, each expected to generate 80 megawatts. Unlike many U.S. and European projects, X-Energy’s design relies on helium gas flowing through 200,000 billiard ball-sized fuel pebbles to capture heat for steam turbines. Planned deployments include 300 megawatts of capacity in the Pacific Northwest and Virginia.

As AI’s hunger for power reshapes energy investment, Big Tech’s bet on next-generation nuclear marks a significant, forward-leaning shift in how America’s computing backbone will be powered in the years ahead.

We need reliable power around the clock to meet the demands of AI.