Artificial intelligence is no longer just a race for better models and faster chips—it is now a battle for electricity. Meta has made that reality unmistakably clear by securing access to as much as 6.6 gigawatts of nuclear power, one of the largest clean-energy commitments ever made by a technology company. The move anchors Meta’s rapidly expanding AI ambitions, including its flagship Prometheus supercluster in Ohio, while reshaping how hyperscalers think about infrastructure, geopolitics, and long-term energy security.
At the center of the strategy is a trio of partnerships with Vistra, TerraPower, and Oklo. Together, the deals give Meta access to firm, carbon-free baseload power through the mid-2030s—precisely the kind of electricity AI data centers require to operate at scale without interruption.
Meta’s approach blends the old and the new. On one side, the company is backing existing nuclear assets operated by Vistra, committing to long-term power purchase agreements that keep legacy reactors online in Ohio and Pennsylvania. Facilities such as Perry, Davis-Besse, and Beaver Valley will not only remain operational, but will also receive funding for capacity upgrades, adding more than 400 megawatts to an already strained regional grid. In doing so, Meta avoids the congestion and delays of traditional grid interconnection queues while ensuring its energy demand does not shift costs onto local ratepayers.
On the other side of the equation, Meta is investing in next-generation nuclear technology. Through agreements with TerraPower and Oklo, the company is effectively acting as a venture-style backer for small modular reactors (SMRs) that promise faster deployment, factory-built components, and lower long-term costs. TerraPower’s Natrium reactors and Oklo’s Aurora fast reactors are still years from full commercial rollout, but Meta’s early financial commitments help move them from concept to construction.
This dual-track strategy is deliberate. Existing reactors provide immediate, dependable power for today’s data centers, while advanced reactors position Meta for the explosive growth expected later in the decade. Together, they could supply up to 6.6 GW of nuclear capacity by 2035—enough to power multiple AI “intelligence factories” operating around the clock.
The stakes are high. AI workloads demand enormous amounts of electricity, and unlike traditional cloud computing, they cannot easily throttle down during peak demand. For Meta, nuclear power offers a rare combination of scale, reliability, and zero-carbon output. Wind and solar remain part of the broader energy mix, but they cannot provide the continuous, always-on power required by massive AI clusters without extensive—and expensive—storage.
Beyond infrastructure, the move has broader implications. By becoming one of the largest corporate buyers of nuclear power in the United States, Meta is helping revive an industry long hampered by high costs and regulatory hurdles. Its deals send a powerful signal to markets and policymakers alike: nuclear energy is no longer just a public utility concern, but a strategic asset for the digital economy.
The decision also positions Meta ahead of rivals in a rapidly intensifying race. As AI capabilities scale, so too will competition for clean electricity. Companies that fail to secure long-term power may find themselves constrained not by algorithms, but by megawatts. In that sense, Meta’s nuclear push is as much about competitive advantage as sustainability.
Prometheus, Meta’s massive AI supercluster in New Albany, Ohio, stands as the clearest beneficiary of this strategy. Designed to train and run next-generation AI systems, the facility requires unprecedented energy density. Nuclear power ensures it can operate continuously without adding carbon emissions or destabilizing local grids.
In practical terms, Meta is redefining what it means to build AI infrastructure. No longer content with buying renewable credits or offsetting emissions after the fact, the company is underwriting real generation capacity. It is paying for power at the source, committing capital decades in advance, and assuming responsibility for its energy footprint.
As AI continues to scale, Meta’s move may prove to be a turning point for the entire tech sector. The future of artificial intelligence, it seems, will be written not just in code—but in concrete, steel, and nuclear fuel.
