The race to dominate artificial intelligence has entered a new phase—and it's being powered by nuclear reactors. Meta has emerged as one of the most significant corporate purchasers of nuclear energy in American history, announcing a sweeping set of agreements that will deliver up to 6.6 gigawatts of power to fuel its AI ambitions. The scale is staggering: that's enough electricity to power approximately five million homes.

A Nuclear Trifecta

Meta's nuclear strategy isn't a single bet but a diversified portfolio spanning three distinct companies and technologies. The approach reflects both the urgency of AI power demands and the uncertainty around which nuclear solutions will prove most scalable.

Vistra: Powering Up Existing Plants

The most immediate component of Meta's nuclear push involves Vistra, one of the largest power producers in the United States. Meta has signed 20-year agreements to purchase electricity from three Vistra-owned nuclear plants: the Perry and Davis-Besse facilities in Ohio, and the Beaver Valley plant in Pennsylvania.

But this isn't just a power purchase agreement—it's an investment in capacity expansion. As part of the deal, Vistra will upgrade these plants to increase energy production, with an additional 433 megawatts of capacity scheduled to come online in the early 2030s. Meta is essentially extending the operational lifespan of these facilities while boosting their output.

Oklo: The Advanced Reactor Bet

The most futuristic element of Meta's strategy involves Oklo, a company developing compact advanced nuclear reactors. Under the agreement, Oklo will deliver up to 1.2 gigawatts of capacity from reactors planned for Ohio, with the first unit potentially operational as early as 2030.

Oklo's Aurora Powerhouse reactors each produce 75 megawatts of electricity—far smaller than traditional nuclear plants but designed for faster deployment and greater flexibility. To fulfill Meta's order, Oklo will need to build more than a dozen of these compact units.

"These partnerships make Meta one of the most significant corporate purchasers of nuclear energy in American history."

— Meta corporate announcement

The Oklo connection carries an interesting Silicon Valley subplot: OpenAI's Sam Altman is one of Oklo's largest investors, owning a 4.3% stake worth approximately $650 million. The company went public in 2024 through a special purpose acquisition company that Altman co-founded. The AI power race is creating some unusual bedfellows.

TerraPower: Bill Gates' Nuclear Vision

The third leg of Meta's nuclear stool involves TerraPower, the advanced nuclear company backed by Bill Gates. Meta has agreed to support the development of two TerraPower reactors capable of generating up to 690 megawatts, with delivery expected as early as 2032.

TerraPower's Natrium reactor design represents a different approach to advanced nuclear—larger than Oklo's compact units but incorporating innovative cooling technology that promises improved safety and efficiency.

The Prometheus Supercluster

All this nuclear power has a specific destination: Meta's Prometheus supercluster computing system being built at a data center in New Albany, Ohio. Expected to come online sometime in 2026, Prometheus represents Meta's bet on AI infrastructure dominance.

The scale of modern AI systems demands energy resources that would have seemed fantastical just years ago. Training large language models requires enormous computational power sustained over weeks or months. Running inference—the process of actually using these models—multiplies energy demands as AI applications scale to billions of users.

The Big Tech Nuclear Rush

Meta isn't alone in turning to nuclear power for AI ambitions. A remarkable convergence is underway as the technology industry's largest players all pursue similar strategies:

  • Google: Signed a deal with Kairos Power to purchase electricity from advanced reactors for its data centers
  • Amazon: Partnered with Energy Northwest and Dominion Energy to develop nuclear capacity
  • Microsoft: Committed to a 20-year deal to restart Pennsylvania's Three Mile Island plant

The pattern is unmistakable: the companies that will shape the AI future are betting heavily on nuclear as the only realistic path to the vast, reliable, carbon-free power their ambitions require.

Market Reaction: Nuclear Stocks Surge

Wall Street has taken notice. Following Meta's announcement, Vistra shares jumped as much as 16%, while Oklo surged as much as 19%. The moves reflect growing investor conviction that nuclear energy represents a durable theme rather than a speculative play.

For investors, the Meta deals validate the thesis that AI's power demands will create sustained opportunities in nuclear energy. Companies positioned to serve this market—whether through existing plant operations, advanced reactor development, or supporting services—may benefit from what could be a multi-decade infrastructure buildout.

The Timeline Challenge

The one asterisk on Meta's nuclear ambitions involves timing. While the Vistra agreements provide near-term power from existing plants, the Oklo and TerraPower components won't deliver meaningful capacity until 2030 or later.

This timeline mismatch highlights a broader challenge facing the AI industry: computing power demands are growing faster than clean energy infrastructure can be built. The current generation of AI systems is being powered largely by existing grid resources, with all the carbon implications that entails.

The Regulatory Factor

Advanced nuclear faces regulatory hurdles that could affect these timelines. While the Nuclear Regulatory Commission has made progress in establishing licensing pathways for new reactor designs, approvals remain time-consuming and uncertain. Any delays could push the promised capacity further into the future.

What This Means for the Energy Transition

Meta's nuclear push carries implications beyond the AI industry. By committing billions of dollars to nuclear capacity—both existing and advanced—Big Tech is providing the demand signal that the nuclear industry has long needed to justify investment in new capacity.

If these projects succeed, they could demonstrate the viability of advanced nuclear designs that have languished in development for years. The financing, operational experience, and public acceptance generated by tech-backed projects could accelerate nuclear deployment across the broader economy.

The Bottom Line

Meta's 6.6-gigawatt nuclear commitment represents more than a corporate energy strategy—it's a statement about what AI's future will require. The company is betting that artificial intelligence will be so central to its business that securing dedicated, reliable, carbon-free power is worth billions in long-term commitments.

For investors, the message is clear: following the AI theme means following the power. The companies enabling AI's energy consumption may prove as important as the companies building AI models themselves. Meta has placed its nuclear bets. The coming years will reveal whether the gamble pays off.