Meta Builds a Nuclear Supply Chain for the AI Era

Meta’s power announcements in January aren’t a simple case of “Meta goes nuclear.” They are better understood as Meta assembling a nuclear supply chain, using three different deal structures to target three different bottlenecks: near-term firm power, medium-term life extension and uprates at existing plants, and longer-term new-build advanced reactors.

Meta says the combined package could support up to 6.6 gigawatts (GW) of new and existing clean power by 2035, building on its earlier nuclear offtake agreement with Constellation Energy and folding these moves into its broader push to scale AI and data center infrastructure.

Part 1: A 20-Year Offtake Tied to Operating Reactors (Vistra)

Meta’s agreement with Vistra isn't a flashy “new reactor” announcement. It is something more important for the next decade of AI-era power: a long-duration financial commitment designed to keep existing nuclear plants running, push more megawatts (MW) out of them, and justify another round of 20-year license extensions. This is happening inside the tightest, most politically contentious power market in the U.S. right now: PJM, the Pennsylvania-New Jersey-Maryland Interconnection, currently the largest Regional Transmission Organization in the country.

The agreed-upon number is a big one: 20-year power purchase agreements covering more than 2,600 megawatts of zero-carbon nuclear energy tied to three Vistra plants: Perry (Ohio), Davis-Besse (Ohio), and Beaver Valley (Pennsylvania). A meaningful share of that commitment is expected to come from uprates, or capacity increases, rather than simply reallocating existing output.

The implication is straightforward. By making this commitment, nuclear power moves from at-risk legacy baseload into foundational power for AI-era infrastructure. Meta is effectively acting as a long-term anchor tenant, similar to how hyperscalers once treated early renewables to catalyze that market; but adapted to a reality where wind and solar alone cannot support 24/7 load growth.

This is the fastest path to firm, carbon-free power for AI workloads. It also turns hyperscaler demand into a bankable revenue backstop for uprates, refueling-cycle economics, and capital spending tied to license extensions. This is exactly the kind of “boring” work that will keep today’s nuclear fleet alive long enough for the SMR era to arrive.

Part 2: The Next Piece — TerraPower’s Natrium as a Scalable Option Set (Up to Eight Plants)

In parallel, Meta struck an agreement with TerraPower to support deployment of up to eight Natrium reactor-and-energy-storage plants in the U.S. TerraPower says this fleet could provide Meta up to 2.8 GW of baseload power; or as much as 4 GW when Natrium’s integrated energy-storage system is used to boost output during peak demand.

Meta’s own summary emphasizes the first step: funding support for two Natrium units totaling up to 690 MW, with delivery targeted as early as 2032, plus rights to energy from up to six additional units (about 2.1 GW) targeted by 2035. Meta is not just buying future power; it is investing to move Natrium from a prototype into a repeatable product. For AI infrastructure planners, the real differentiator is not reactor chemistry; it is repeatability and schedule confidence across multiple sites.

The agreement is meaningful, but it is still in the site-selection and early-development phase, which is often where large infrastructure projects either gain momentum or stall out. For now, the companies say they expect to identify a specific site for the initial two-reactor project “in the coming months,” a milestone that will determine whether Natrium shifts from promise to pipeline.

Part 3: The “Build Me a Power Campus” Piece — Oklo and a Prepay Mechanism in Ohio (1.2 GW)

Meta’s agreement with advanced-reactor developer Oklo to support a scalable, up-to-1.2-gigawatt nuclear power campus in Pike County, Ohio is one of the clearest signals that hyperscalers are moving beyond “buy clean credits” and into direct, balance-sheet-backed development of new firm power. And it is aimed squarely at feeding AI-era load growth.

The standout feature of the deal is not just the headline number. It is the structure: Meta can prepay for power and provide early funding to de-risk fuel procurement and early project work. This goes straight at the two choke points that have slowed many first-of-a-kind advanced nuclear projects. Prepayment is a financing innovation as much as an energy deal: hyperscaler-style capex thinking applied to power. Cash is pulled forward to de-risk early stages like site work, interconnection studies, and supply-chain reservations, in exchange for a clearer line of sight to future firm power.

The Meta deal is getting the headlines, but Oklo’s most important near-term progress may be its deepening integration with DOE fast-track frameworks, especially around fuel.

Oklo reported that DOE’s Idaho Operations Office approved a Nuclear Safety Design Agreement for the Aurora Fuel Fabrication Facility at Idaho National Laboratory, the first such agreement under DOE’s Fuel Line Pilot Projects, designed to demonstrate a faster authorization pathway.

DOE’s Fuel Line Pilot Program is meant to support the Reactor Pilot Program and establish domestic fuel production lines using DOE authorization processes. This speaks to the real bottleneck facing advanced nuclear: development is increasingly constrained not by reactor design, but by HALEU availability, fuel-form qualification, and fabrication throughput. Fuel readiness is what turns a “reactor on paper” into a reactor you can actually start; which brings us to…

Part 4: DOE’s Parallel “Nuclear Renaissance” Push — Uranium Enrichment and the AI Load Problem Converge

DOE’s nuclear posture right now is not a single program. It is a stack that runs from fuel supply to federal land reuse to planning for AI-driven electricity demand.

On January 5, 2026, DOE announced the availability of $2.7 billion to strengthen domestic uranium enrichment services over the next decade. The announcement explicitly includes support for HALEU-related supply-chain innovations.

Reuters added detail, reporting awards to American Centrifuge Operating, General Matter, and Orano Federal Services, framed around both reducing reliance on Russian supply and supporting HALEU needs for advanced reactors. DOE is effectively acknowledging that AI-era load growth and advanced nuclear deployment share a hard dependency: a domestic enriched-fuel industrial base.

At the same time, DOE’s Office of Environmental Management issued a solicitation seeking proposals to build and power AI data centers on DOE’s Paducah site in Kentucky. The site is being positioned as part of a broader effort to reuse federal land for energy and data center development, with private applicants responsible for funding, building, operating, and ultimately decommissioning facilities.

Taken together, these moves show DOE treating fuel, land, and load growth as part of the same problem set: how to build a nuclear and energy system that can actually keep up with the scale and speed of AI infrastructure.

What Really Matters Next — and What to Watch Closely

Meta’s nuclear strategy sketches a long-arc vision for powering AI infrastructure. But between today’s announcements and tomorrow’s reactors sits a messy middle defined by timing gaps, fuel constraints, regulatory risk, and rising political scrutiny. These are the pressure points that will determine whether this strategy becomes a template or a cautionary tale.

Power development timelines are long; AI timelines are not

Meta’s Prometheus AI data center in Ohio is slated to come online in 2026, while Natrium deliveries are discussed in the 2032–2035 window. That gap means Meta still needs bridging strategies: existing nuclear offtake, gas paired with offsets, grid upgrades, storage, and new regional transmission. The nuclear stack is the destination; but for now, it is only part of the journey.

Fuel supply and licensing remain decisive

DOE is pushing fuel and enrichment, but scaling HALEU is an industrial challenge with geopolitical constraints. The industry will need to watch for real signals of progress: construction starts, NRC filings, fuel-qualification milestones, and whether DOE authorization pathways translate into repeatable commercial licensing outcomes.

The “data center as a grid actor” backlash is growing

States are increasingly scrutinizing how large loads affect ratepayers and grid costs. New York’s recent moves are one example, along with new large-load frameworks in Virginia and Georgia. As nuclear-backed AI campuses scale, expect more pressure on cost allocation, interconnection queues, and the question that will not go away: who pays for transmission?

When Politics Catches Up to Power: The PJM Test Case

The politics of power are now catching up to the economics of AI.

As of mid-January, the Trump administration has urged PJM to consider an emergency electricity auction that would force large power users, explicitly including tech companies and data centers, to commit to long-term contracts that would fund new generation. The rationale is straightforward: surging AI and crypto demand is stressing capacity markets, pushing up consumer bills, and exposing how fragile the current planning model has become.

Whether or not PJM ultimately adopts that approach, the signal is clear. Policymakers are increasingly unwilling to let massive new loads ride on legacy grid assumptions. The question is no longer just how fast data centers can connect; it is who pays to make the system bigger, stronger, and more reliable.

Seen through that lens, Meta’s nuclear strategy looks less like a branding exercise and more like a hedge against a future in which “pay your way” becomes formal policy. Long-term nuclear offtake, prepay mechanisms, and direct development of new power aren’t just about decarbonization. They are about control—over cost, over timelines, and over political risk.

If emergency auctions, special rate classes, or federal interventions become the norm, hyperscalers will face a choice: absorb unpredictable grid politics, or build their own path to firm power. Meta is betting that nuclear, old and new, can be that path. Whether the grid, regulators, and fuel supply chain can keep pace will determine whether this strategy becomes the industry’s template, or simply its most ambitious experiment.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.