Meta Platforms has locked in 20-year power purchase agreements with Vistra Corp. to secure more than 2,600 megawatts of zero-carbon nuclear energy for its expanding artificial intelligence data centers. The contracts cover output from three nuclear plants operating within the PJM regional grid and include provisions for planned capacity expansions at those facilities. With Meta’s Prometheus AI data center project already taking shape in New Albany, Ohio, the deal signals how aggressively Big Tech is moving to guarantee reliable, carbon-free electricity for compute-intensive AI workloads.
Why Meta’s 6.6‑gigawatt nuclear bet matters now
The scale of this commitment sets it apart from earlier corporate clean‑energy purchases. Vistra disclosed the agreements in a Form 8‑K filed with the SEC, confirming 20‑year PPAs with Meta. That term length alone is unusual for power contracts tied to a single corporate buyer and reflects the capital certainty both companies need to justify billions in infrastructure spending. For Meta, it is a way to lock in long‑duration, carbon‑free power in a market where AI‑driven demand is rising faster than new generation is coming online.
The three plants covered by the deal are Beaver Valley in Pennsylvania, Davis‑Besse in Ohio, and Perry, also in Ohio. All three sit inside the PJM Interconnection, the grid operator serving 13 states and the District of Columbia. PJM has faced tightening reserve margins as data‑center demand climbs across Virginia, Ohio, and surrounding states. A long‑term offtake agreement of this size gives Vistra a financial anchor to keep those reactors running and to pursue output increases that would add new generation to a grid under growing strain.
Meta’s reported goal of ultimately accessing roughly 6.6 gigawatts of nuclear capacity through these arrangements, once uprates and expansions are completed, underscores how central AI has become to its long‑term planning. Unlike intermittent wind and solar, nuclear plants can run at high capacity factors, providing the around‑the‑clock power profile that large language models, recommendation engines, and other AI systems demand. In that sense, the nuclear contracts are not just a climate statement but a core piece of Meta’s compute strategy.
The central policy question is whether these contracts will speed up the regulatory process for expanding plant output. The Nuclear Regulatory Commission has already reviewed and approved power uprate applications for Beaver Valley Units 1 and 2. If Vistra files uprate applications for Davis‑Besse and Perry on a similar timeline, observers will be watching NRC docket activity over the next 18 months to see whether a deep‑pocketed corporate backer changes the pace of approvals compared with prior non‑corporate‑backed filings. Any acceleration would raise questions about consistency in how uprates are evaluated across different plant owners and customers.
The evidence behind Meta’s nuclear power contracts
Vistra’s own announcement described the agreements as providing more than 2,600 MW of zero‑carbon energy from the three named plants. In that press release, the company framed the contracts as supporting both continued operation of existing reactors and planned expansions and uprates. That dual purpose matters because keeping aging nuclear plants online requires sustained capital investment in maintenance, fuel, and staffing. Without a guaranteed buyer for output at a predictable price, plant owners face the risk that wholesale electricity prices will fall below operating costs, as happened to several U.S. reactors that shut down prematurely over the past decade.
The Associated Press separately reported that Meta’s agreements are connected to its Prometheus AI data center project in New Albany, Ohio. That geographic proximity is significant. Davis‑Besse and Perry both operate in Ohio, meaning the electrons generated by those plants flow into the same regional grid that will serve the Prometheus campus. While power purchase agreements do not guarantee that specific electrons travel from a reactor to a data center, matching generation and consumption within the same grid region reduces transmission losses and strengthens the environmental accounting behind carbon‑free claims.
The headline figure of 6.6 gigawatts reflects the total nuclear commitment once uprates and expansions are complete. No single primary document published so far breaks down the exact megawatt allocation plant by plant or specifies the precise capacity additions expected from each uprate. The SEC filing confirms the contractual structure and counterparties. The Vistra press release confirms the aggregate capacity exceeds 2,600 MW from the three plants in their current configuration. The gap between 2,600 MW of existing output and the 6.6‑GW total suggests that planned expansions and new generation could roughly double or triple the contracted capacity over the life of the agreements, though the specific engineering and licensing steps required to reach that figure have not been detailed in public filings.
For now, the public record supports three core points: Meta has agreed to long‑term PPAs with Vistra; the deals cover three named nuclear plants in PJM; and the initial contracted capacity is more than 2,600 MW, with an expectation of substantial growth as uprates and expansions proceed. Anything beyond that, including the precise 6.6‑GW configuration, remains an informed inference rather than a fully documented fact.
What remains unresolved for Meta’s nuclear strategy
Several open questions will shape whether this deal delivers on its promise. First, Meta has not issued its own detailed statement explaining how the Vistra power will be distributed across its data‑center portfolio. The Prometheus project in Ohio is one known destination, but Meta operates and plans facilities in multiple states. Without that allocation map, it is difficult for outside analysts to assess whether the nuclear supply will be concentrated at a few flagship AI campuses or spread thin across many sites.
Second, the NRC records for Beaver Valley confirm that uprate reviews have been completed for Units 1 and 2, but publicly available docket information does not yet tie specific approved capacity figures to the Meta PPAs. For Davis‑Besse and Perry, Vistra has indicated an intention to pursue additional output, but has not, in the documents released so far, laid out a schedule for submitting uprate applications or described the scope of the engineering modifications required. That leaves uncertainty about when, and in what increments, the contracted capacity will actually rise.
Third, there is the question of grid integration. PJM has warned that rapid load growth from data centers, electrification, and industrial projects could outpace new generation and transmission unless investments accelerate. While nuclear plants provide dependable baseload power, they do not on their own solve local transmission bottlenecks or substation constraints near fast‑growing data‑center clusters. Meta’s long‑term contracts help keep generation available on the system, but the company will still depend on PJM planning processes and utility‑led upgrades to move that power to where it is needed most.
Finally, the broader policy debate over nuclear’s role in the energy transition will continue to shape public perception of deals like this one. Supporters argue that pairing AI with nuclear power avoids locking in fossil‑fuel generation to serve new digital demand and helps preserve jobs at existing plants. Critics point to unresolved issues around nuclear waste, safety, and cost, and question whether such large corporate‑backed contracts could crowd out investment in renewables or grid‑scale storage. With limited details about pricing and risk‑sharing mechanisms in the Meta–Vistra agreements, those debates will likely play out in state regulatory arenas and public forums rather than in the contract documents themselves.
What is clear is that Meta’s decision to underwrite decades of nuclear output marks a turning point in how technology companies think about energy. Rather than relying solely on short‑term renewable energy credits or scattered solar and wind deals, the company is aligning its AI ambitions with long‑lived, capital‑intensive infrastructure. As more data‑center operators confront similar growth curves, the Meta–Vistra model-anchoring existing nuclear fleets while financing uprates and new capacity-could become a template for how digital and power‑sector giants negotiate the next phase of the clean‑energy transition.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
