Meta announced a cluster of long-term nuclear deals on Jan. 9 that could add as much as 6.6 gigawatts of clean, firm power to U.S. grids by 2035 — a move the company says is aimed squarely at powering its surging AI data-center needs.
The social‑media giant signed 20‑year agreements covering three operating Vistra nuclear plants, and struck development partnerships with TerraPower and Oklo to support advanced and small modular reactor projects. Meta frames the package as both an energy-security play for its Prometheus supercluster in Ohio and an investment in U.S. nuclear supply chains and jobs. Meta’s full announcement is posted on the Meta newsroom.
What the deals include
- Vistra: Long‑term power purchase agreements for more than 2,176 MW from Vistra’s Perry and Davis‑Besse plants in Ohio, plus support for uprates that add roughly 433 MW across Perry, Davis‑Besse and Beaver Valley in Pennsylvania — bringing the Vistra‑backed portion to about 2,609 MW. Those PPAs are intended to help finance plant upgrades and license‑extension plans that keep reactors online for decades.
- TerraPower: Funding and rights tied to Natrium reactor deployments. Meta will support two Natrium units capable of delivering up to 690 MW as early as 2032, and holds options for energy from additional Natrium units that could produce roughly 2.1 GW by 2035. TerraPower says the agreement helps speed commercial deployment of its advanced design.
- Oklo: A development partnership aimed at bringing as much as 1.2 GW of new advanced nuclear capacity to Pike County, Ohio, potentially online by about 2030. Meta’s support is described as early procurement and development funding to help Oklo move construction readiness along.
Together with a prior 20‑year deal Meta signed with Constellation to keep an Illinois reactor operating, the new agreements are intended to supply up to 6.6 GW of new and existing nuclear capacity across the coming decade.
Why Meta is doing this now
AI workloads are voracious. Meta’s Prometheus cluster in New Albany, Ohio — announced last year as a 1‑GW regional supercluster — and the company’s broader expansion of high‑density compute require predictable, always‑available electricity. Nuclear provides baseload — steady, carbon‑free power — that complements intermittent wind and solar as data centers scale.
This strategy mirrors broader industry thinking about how to meet rising AI demand without destabilizing grids: lock in long‑term, firm generation and finance new capacity well ahead of load. For context on other bold power ideas tied to AI centers, see how companies are exploring off‑earth options like Google’s Project Suncatcher. And as cloud models and search tools get more compute‑heavy — illustrated by initiatives like Gemini Deep Research — the appetite for reliable power only grows.
Local economies and the nuclear supply chain
Vistra’s statements on the deals highlight job preservation and creation: uprate projects will require thousands of project‑period construction roles and help sustain the thousands of permanent jobs at the plants. Meta and Vistra both position the contracts as protecting local tax bases and encouraging investment in aging but capable nuclear assets.
Meta argues its commitments also shore up the nascent U.S. advanced nuclear industry by helping TerraPower and Oklo reach development milestones and attract additional capital. TerraPower says the deal provides market certainty for Natrium units; Oklo frames Meta’s commitment as a major step toward delivering multiple Aurora reactors at scale.
Realistic limits and open questions
This is not all upside. Oklo’s and TerraPower’s reactors are advanced designs that still face regulatory review, permitting and the hard economics of novel builds; there are currently no U.S. commercial SMRs operating at scale. Critics note that factory‑built SMRs must prove cost‑competitiveness and supply‑chain maturity to meet optimistic timetables. Even uprates at existing reactors require careful engineering, licensing and multi‑year workstreams.
From a market perspective, investors reacted quickly: Oklo shares jumped, and Vistra rallied on the news, reflecting market confidence in financing paths when large corporate buyers step in.
Bigger picture
What Meta is doing is more than hedging a bill for kilowatt‑hours. It’s an example of a tech company using procurement—and capital commitments—to de‑risk emerging energy projects and shape the grid it depends on. Whether that accelerates a broader wave of corporate‑backed nuclear projects will depend on execution: successful uprates, timely NRC approvals, and whether advanced reactors can move from demonstration to reliable commercial builds.
Meta’s playbook — long deals, development funding, and a clear tie to specific data‑center projects — may become a template for other hyperscalers that face identical grid‑integration challenges. If it works, the reward is steadier electricity and a smoother path for the next generation of compute. If it stumbles, the industry will have learned how hard it still is to build nuclear at scale in the 2020s.
Meta’s announcements are a vivid reminder that the future of AI and the future of power are intertwined. The next few years — licensing, construction starts, and the first uprate completions — will show whether this marriage of silicon and uranium was prescient or premature.