Morning Overview

Meta signed one of the largest corporate nuclear deals in US history to power its AI.

Meta Platforms secured 2,609 megawatts of nuclear power through 20-year agreements with Vistra Corp., a deal announced on Jan. 9, 2026, that ranks among the largest corporate nuclear procurement commitments in United States history. The power will feed the Prometheus AI data center project in New Albany, Ohio, as the company races to lock down round-the-clock electricity for its expanding artificial intelligence operations.

Why a 20-year nuclear commitment changes the calculus for AI power

The sheer scale of this deal signals a structural shift in how technology companies plan to fuel data centers. At 2,609 MW, the contracted capacity is roughly equivalent to the output of two and a half large nuclear plants. That volume, locked in for two decades, goes well beyond the shorter-term renewable energy contracts that have dominated corporate clean-power procurement over the past decade. Solar and wind farms can be built quickly, but they generate electricity intermittently. Nuclear reactors produce steady, carbon-free output around the clock, which is exactly the load profile that AI training clusters demand.

The deal also tests a broader hypothesis: that long-duration nuclear offtake agreements from a single buyer can create a procurement track that moves faster than traditional utility-led nuclear construction. Utilities typically plan new nuclear capacity through lengthy regulatory proceedings and rate cases that stretch across years. A direct 20-year power purchase agreement, or PPA, between a generator and a corporate buyer bypasses much of that process. If other hyperscale cloud and AI companies follow Meta’s lead with similar 15- to 20-year commitments, the resulting demand signal could pull forward reactor uprates and restarts that utilities alone would not pursue on the same timeline.

That dynamic matters because AI electricity consumption is growing faster than grid operators anticipated even two years ago. Data center developers across the Midwest and Mid-Atlantic have been filing interconnection requests at a pace that has strained regional planning queues. A dedicated nuclear supply contract lets Meta sidestep some of that congestion by tying its load directly to existing and expanded generation rather than waiting for new transmission or generation to clear the queue.

What the Vistra filing and Meta’s Prometheus project reveal

The primary record of the transaction is Vistra’s SEC filing, which discloses that the company entered into 20-year PPAs with Meta Platforms. The filing specifies a total contracted capacity of 2,609 MW, a figure that includes both existing operating output from Vistra’s nuclear fleet and planned uprates at those facilities. The inclusion of uprates is significant: it means a portion of the megawatts Meta will receive does not yet exist and depends on engineering work to squeeze more generation from reactors already in service.

Vistra operates one of the largest competitive nuclear fleets in the country, with plants in Illinois and Texas. The 8-K does not break down how much of the 2,609 MW comes from current production versus future uprates, and pricing terms, escalation clauses, and termination provisions are redacted in the public filing. Those gaps leave open questions about the financial risk each party is absorbing over two decades, as well as how costs and benefits might shift if wholesale power prices move sharply up or down.

On the buyer side, Meta tied the agreements to its Prometheus data center campus in New Albany, Ohio. The Prometheus campus is designed to support large-scale AI model training and inference workloads, the kind of computing that runs servers at near-full capacity for weeks or months at a time. That usage pattern aligns with nuclear’s always-on generation profile far better than it aligns with variable renewables, which is a practical reason Meta chose this route rather than simply adding more wind and solar PPAs to its portfolio.

The announcement date itself, Jan. 9, 2026, places the deal at a moment when several other technology firms have been publicly exploring nuclear options. Microsoft has pursued arrangements to support a reactor restart, Amazon Web Services has invested in small modular reactor development, and Google has signed agreements tied to advanced reactor technology. Meta’s Vistra deal differs from those efforts in a key respect: it draws power from conventional, already-operating light-water reactors and their planned uprates, not from next-generation designs that remain years from commercial operation. That choice lowers technology risk but concentrates the bet on regulatory and operational execution at existing plants.

Open questions on pricing, grid impact, and whether rivals will match the bet

Several material details remain unresolved. The redacted pricing in Vistra’s SEC filing means outside analysts cannot yet determine whether Meta is paying a premium over prevailing wholesale electricity rates or locking in a discount against expected future price increases. The answer matters because it will influence whether other large buyers view similar contracts as financially attractive or as a costly insurance policy against power scarcity. If the contract price sits well above projected market averages, rivals may hesitate to follow; if it undercuts expected future costs, the deal could be seen as a savvy hedge that others rush to replicate.

The split between existing output and planned uprates also raises execution risk. Nuclear uprates require approval from the Nuclear Regulatory Commission, and the timeline for engineering reviews, safety analyses, and construction can stretch beyond initial estimates. If uprate projects at Vistra’s plants are delayed, Meta could face a period where contracted capacity exceeds delivered megawatts, depending on how the PPA handles shortfalls. The contract could obligate Vistra to procure replacement power on the market, or it might allow temporary reductions in deliveries; from the redacted public record, it is not yet clear how that risk is allocated.

Grid-level consequences are another blind spot. Ohio regulators and PJM Interconnection, the regional grid operator covering New Albany, will have to determine how to treat a large, nuclear-backed data center load that is effectively pre-served by specific generating units. On one hand, tying a power-hungry AI campus to a firm, carbon-free source could ease concerns about reliability and emissions. On the other, a long-term bilateral contract of this size may complicate capacity market planning if a substantial share of a generator’s output is effectively removed from the pool available to serve other customers.

There are also questions about how this arrangement will interact with transmission planning. Even if the nuclear megawatts exist on paper, they still have to flow across the grid to reach New Albany. If the data center’s demand ramps up faster than local transmission upgrades, congestion could emerge, raising costs for Meta or for other customers in the region. Regulators will be watching to see whether the deal prompts new infrastructure proposals and how those costs are allocated.

Whether Meta’s rivals will match this bet is uncertain. Some technology companies may prefer to maintain flexibility rather than locking into 20-year baseload contracts, especially in a sector where computing architectures and energy technologies evolve quickly. Others may lack access to a competitive nuclear fleet willing to sign corporate PPAs on this scale. Still, the visibility of Meta’s move, and the sheer size of the commitment, effectively sets a new benchmark for what it means to secure “24/7 clean power” for AI workloads.

For now, the Vistra-Meta agreement stands as a test case. If the uprates proceed smoothly, the power flows reliably, and the economics prove favorable, it could validate nuclear PPAs as a mainstream tool for decarbonizing and de-risking data center growth. If delays, cost overruns, or regulatory friction undermine the deal’s performance, it may reinforce arguments that long-term nuclear procurement remains better suited to utilities than to individual corporate buyers. Either way, the contract underscores how the race to build ever-larger AI systems is beginning to reshape the energy system itself, pushing companies to make bets measured not just in megawatts, but in decades.

More from Morning Overview

*This article was researched with the help of AI, with human editors creating the final content.