Meta is turning to nuclear power on a scale rarely seen in corporate America, locking in long-term energy to feed its rapidly expanding artificial intelligence infrastructure. The company has signed one of the largest private nuclear procurement packages in United States history, tying its AI ambitions directly to next-generation reactors and fuel supply. The move signals a new phase in the energy race around AI, where clean baseload power is becoming as strategic as chips or algorithms.
What changed in Meta’s nuclear strategy for AI-scale power
Meta has committed to a suite of nuclear projects that together target 6.6 gigawatts of capacity for its United States data centers, according to an energy procurement plan. That volume rivals the output of several large conventional nuclear plants and would make Meta one of the biggest corporate buyers of nuclear power in the country. The deals are structured around both existing reactors and advanced designs that aim to come online over the next decade.
The company has lined up partnerships with Vistra, TerraPower and Oklo, three players that cover different points on the nuclear spectrum. Vistra controls large operating nuclear assets and development sites that can deliver grid-scale power relatively quickly. TerraPower and Oklo are working on advanced reactors that use novel designs and fuel cycles, and Meta has signaled interest in tying future data center growth to these next-generation projects. Taken together, the agreements shift Meta from a buyer of generic grid power to a direct sponsor of specific nuclear facilities.
One of the most unusual elements is Meta’s willingness to contract directly for nuclear fuel from a startup supplier. The company has agreed to purchase fuel for advanced reactors in a structure that gives the startup predictable demand and gives Meta visibility into a key part of the supply chain for its AI data centers. That kind of fuel offtake is more typical of utilities or governments than of an internet platform, and it shows how deeply Meta is now embedded in nuclear project economics.
Meta is pairing these nuclear commitments with a broader clean energy portfolio that also includes experimental sources. The company is backing space-based solar power that beams electricity from orbit to receivers on the ground, testing whether this technology can complement nuclear as another 24/7 resource for its AI machines. In Meta’s strategy, nuclear is not replacing renewables; it is being layered on top of them to secure round-the-clock capacity.
Internally, this marks a shift from a model where data centers were largely powered by a mix of wind and solar projects, often located far from the facilities themselves. For AI, Meta is now tying energy procurement more directly to specific campuses and clusters, including an AI supercluster that will rely on dedicated nuclear output, according to an industry analysis. The company is signaling that proximity, reliability and regulatory certainty now matter as much as headline renewable percentages.
Why Meta’s giant nuclear bet matters for AI and clean energy
Scale is the most immediate reason this deal matters. AI training and inference workloads demand enormous, always-on electricity. As Meta builds larger clusters of GPUs and custom accelerators, the company needs power that is not only low carbon but also available every second of the day. Nuclear fits that profile, and the 6.6 gigawatt target shows how large AI operators expect their future demand to be. For comparison, a single gigawatt-class nuclear unit can power hundreds of thousands of homes, so Meta is effectively reserving the output of multiple plants for its own digital infrastructure.
The agreements also reshape the conversation around what “clean energy” means for big tech. Meta has framed the nuclear portfolio as a way to support United States climate goals while meeting AI demand, presenting the reactors as a complement to wind and solar rather than a competitor. The company’s public messaging links nuclear expansion directly to American AI leadership, arguing that low-carbon baseload power is a strategic asset in the global race to build advanced models and services.
For the nuclear sector, Meta’s move provides something it has long lacked: a deep-pocketed, non-utility buyer willing to sign early and for large volumes. The deals with TerraPower and Oklo validate advanced reactor business models that depend on securing anchor customers before projects reach construction. By agreeing to buy power and fuel from these advanced nuclear developers, Meta is helping them attract financing and navigate lengthy regulatory processes.
There are also grid-level implications. As more data centers cluster around specific regions, they risk straining local transmission and pushing up prices for other customers. By tying new data centers to dedicated nuclear generation, Meta is trying to reduce the pressure on existing grids and avoid accusations that AI is cannibalizing community power supplies. Analysts following the company’s data center expansion have noted that nuclear-backed campuses can be sited and planned with grid integration in mind, rather than simply drawing from whatever power is available.
Politically, the deals may accelerate the normalization of nuclear energy within climate and technology policy debates. If one of the world’s largest social platforms is comfortable tying its brand and AI strategy to nuclear, other tech firms may feel they have more cover to do the same. That could influence regulatory timelines, subsidy design and public investment in advanced reactors, especially when framed as supporting national competitiveness in AI.
At the same time, Meta’s bet raises questions about who controls future clean power resources. If hyperscale tech companies sign long-term contracts for large shares of new nuclear capacity, smaller buyers and communities may have less access to firm, low-carbon electricity. The company’s deals highlight a shift from utilities as the primary counterparties for nuclear projects to a model where corporate buyers can shape which plants get built and where.
What comes next for Meta’s nuclear-powered AI buildout
The nuclear projects Meta is backing will take years to fully materialize, so the near term will be defined by development milestones and regulatory decisions. Vistra’s assets and expansion plans can deliver power on a shorter horizon, which is why they are central to Meta’s initial procurement package. TerraPower and Oklo will need to navigate design certification, siting and construction before their reactors can feed Meta’s data centers, and those processes will test how quickly advanced nuclear can scale to meet AI demand.
Meta has also launched a request for proposals tied to the 6.6 gigawatt goal, inviting additional nuclear projects to compete for long-term offtake contracts. The RFP signals that the current deals are a starting point rather than the final word on the company’s nuclear portfolio. Developers that can offer firm, low-carbon power at competitive prices and with credible timelines will see Meta as a marquee customer, which may spur new reactor designs or repowering plans for existing sites that can align with the company’s clean energy.
On the AI side, Meta’s infrastructure roadmap will increasingly be constrained or enabled by how quickly these energy projects move. Large training clusters, inference farms for products like recommendation engines and generative tools, and new services that run continuously all depend on predictable power. If nuclear projects slip, Meta may need to rely more heavily on grid purchases or accelerate other experiments such as the space-based solar tests it has already begun.
Other tech companies are watching closely. If Meta’s approach proves viable, a second wave of hyperscalers could follow with their own nuclear-backed data center plans, potentially bidding against each other for limited advanced reactor capacity. That would give nuclear developers more leverage but could also expose them to concentrated customer risk if AI demand forecasts change. The structure of Meta’s fuel and power will serve as a template for how future corporate buyers engage with the sector.
