La apuesta nuclear de Meta: Asegurar la infraestructura de IA

The core investment thesis here is clear: Meta is building the fundamental rails for the next technological paradigm. Its 6.6 gigawatts of nuclear power by 2035 is not just an energy purchase; it is a deliberate, first-principles infrastructure bet to secure exponential AI growth. This move is driven by a hard, physical constraint: the massive compute power required for projects like the Prometheus supercluster, which is set to come online this year. As AI data centers increase U.S. power demand for the first time in two decades, the company is positioning itself as the leading corporate buyer to lock in supply.

Meta's approach shows a calculated, forward-looking strategy. It didn't just sign a contract; it used a prior request for proposals to solicit interest for 1-4 gigawatts, demonstrating a methodical effort to shape the market. This deal with Vistra is the first major step, but it's part of a broader portfolio that includes partnerships with SMR developers TerraPower and Oklo. By funding the construction of new reactors, Meta is moving beyond a passive buyer to an active developer, aiming to accelerate deployment and build a competitive moat around its AI operations.

The scale is staggering. 6.6 gigawatts is enough energy to power Ireland, a figure that underscores the sheer magnitude of the compute challenge. This isn't about incremental growth; it's about securing the foundational energy layer for an entire S-curve of adoption. Meta's goal is to ensure its AI infrastructure isn't bottlenecked by grid constraints, a vulnerability that could slow its entire technological trajectory. In this race, the company is betting that controlling the power source is as critical as controlling the software.

The Infrastructure Play: Building the Rails for the Next Paradigm

Meta's nuclear portfolio is a masterclass in infrastructure building for an exponential future. It's not a single bet, but a layered strategy designed to secure power across the entire technological S-curve. The deals combine immediate, firm capacity with long-term wagers on next-generation technology, creating a dual-track approach to de-risking the AI energy supply chain.

The structure is deliberate. A core component is 2,176 MW of power from existing Vistra plants, providing reliable, near-term energy to feed its current and near-future data centers. This is the essential first mile. Then comes the forward-looking capital. Meta is funding the development of up to eight Natrium sodium fast reactors from TerraPower, with the first units targeted for delivery as early as 2032. Simultaneously, it is supporting Oklo's 1.2 GW power campus in Ohio, with Meta prepaying for power to advance project certainty. This mix-existing plants for today, advanced reactors for the mid-term-ensures a smooth transition as the AI compute demand curve steepens.

Meta's role here is crucial as a de-risking funder. Advanced nuclear projects face a notorious "valley of death" between prototype and commercial deployment. By providing upfront capital and long-term power purchase agreements, Meta is stepping into that gap. This financial commitment gives developers like TerraPower and Oklo the commercial certainty needed to attract further investment and accelerate construction timelines. In effect, Meta is acting as a first customer and anchor tenant, helping to scale the technology and drive down costs-a classic move to build the rails for an entire industry.

The scale of this infrastructure play is matched by the sheer capital commitment to the AI stack itself. The $27 billion joint venture for the Hyperion data center campus is a massive bet on physical capacity. With an expected computing capacity of five gigawatts, this single project represents a significant portion of the total 6.6 gigawatts Meta is securing. It's a direct investment in the compute layer, ensuring it has the physical space and power to deploy its AI models at an exponential rate. The complex financial structure, with a 16-year operational guarantee, underscores the long-term, capital-intensive nature of this infrastructure build-out.

Together, these moves form a comprehensive infrastructure stack. Meta is not just buying power; it is funding the reactors that will generate it, and building the data centers that will consume it. This integrated approach is the hallmark of a company building the fundamental rails for the next paradigm. It's a strategy designed to outlast the volatility of energy markets and the construction timelines of nuclear projects, securing its position at the center of the AI S-curve for the next decade.

Financial and Execution Risks: The Cost of Being First

The aggressive infrastructure build-out is a massive bet, but it comes with tangible costs. Meta is bearing the full financial weight of its energy and data center strategy, a commitment that introduces significant P&L and execution risks.

First, there is the direct cost of energy. Meta has stated it will pay the full costs for energy used by our data centers, a promise that shifts the entire burden from consumers to the company. While this secures supply, it also means Meta's operating expenses will absorb the full price of nuclear power, including any volatility or premium for securing new capacity. This is a major line item for a company that is already planning to spend "hundreds of billions of dollars" on data centers. The financial impact will be felt in its margins, especially if power prices rise or construction overruns occur.

Then there is the local friction. The Hyperion project's scale is provoking a backlash in its immediate community. Construction has already caused a 600 percent increase in crashes in the nearby town of Holly Ridge, leading to safety concerns for residents and students. Reports of "Meta drivers who 'think they run this road'" and incidents like a hit-and-run by a Meta dump truck highlight the social and regulatory risks of deploying such a massive, resource-intensive facility in a rural area. This local opposition can delay permitting, increase community relations costs, and create a public relations vulnerability that could slow the project's timeline.

Finally, the financial structure of the Hyperion JV locks Meta into a long-term commitment. While the company retains a 20% ownership stake, it has provided the joint venture with a guarantee for the first 16 years of operations. This means Meta would have to make cash payments if leases are not renewed, creating a fixed liability that could limit its financial flexibility. This 16-year guarantee is a significant bet on the sustained, exponential growth of AI compute demand. If that forecast shifts, the company could be on the hook for years of payments for capacity it may not need, a classic risk of being a first-mover in a rapidly evolving infrastructure market.

The bottom line is that Meta's strategy to build the rails for the AI S-curve is capital-intensive and carries real execution costs. The company is paying for the privilege of being first, and those costs manifest in its P&L, its community relations, and its long-term financial obligations.

Catalysts and What to Watch

The investment thesis hinges on execution. For Meta's nuclear and data center bets to pay off, several near-term milestones must be met. The key metrics to watch are the construction timelines for its advanced reactors and the Hyperion campus, the actual growth of its AI compute capacity, and the smooth progression of regulatory and community approvals.

First, track the construction schedules. The company has committed to funding new reactors, with the first units from TerraPower's Natrium project potentially coming online "as early as 2030". The broader portfolio aims for 6.6 gigawatts by 2035. Any significant delays here would directly challenge the thesis that Meta is securing power for its AI S-curve. Similarly, the $27 billion Hyperion data center campus is slated to come online in 2030. Construction is already underway and causing local disruption, so monitoring for cost overruns or permitting hold-ups is critical. These projects are the physical proof points of Meta's infrastructure strategy.

Second, measure Meta's own compute demand against its secured supply. The company's roadmap requires massive power. The Prometheus supercluster, expected to come online this year, will consume at least 1 GW. The Hyperion campus itself has a computing capacity of five gigawatts. The total secured portfolio of 6.6 GW must not only cover these projects but also the exponential growth of future AI workloads. Investors should watch for updates on Meta's AI compute capacity growth and data center power consumption to see if the secured energy is sufficient or if the company will need to make further, costly commitments.

Finally, monitor the regulatory and community front. The Hyperion project is already provoking a backlash in Holly Ridge, Louisiana, with a 600 percent increase in crashes and local safety concerns. This friction is a direct execution risk that could lead to delays or increased costs. For the nuclear reactors, regulatory approvals from the NRC are a critical, multi-year hurdle. Any setbacks in these areas would slow the deployment of the very infrastructure Meta is betting on. The smooth passage of these non-technical approvals is as important as the construction milestones themselves.

The bottom line is that Meta's strategy is a long-term play on the AI infrastructure S-curve. The near-term catalysts are the tangible signs that the company is delivering on its promises. Watch the construction timelines, the power consumption metrics, and the regulatory progress. These are the data points that will validate whether Meta is successfully building the rails for the next paradigm-or facing a costly detour.