Meta Platforms, Inc. (META) has signed long-term agreements for 2.5 gigawatts of nuclear-generated electricity to support its growing AI infrastructure, reinforcing BofA Securities' Buy rating. The deals mark a strategic pivot toward stable, low-carbon energy for data centers.
- Meta secured 2.5 gigawatts of nuclear energy through power purchase agreements
- Agreements involve facilities in Pennsylvania and Georgia, with initial deliveries in 2027
- Energy capacity supports AI training clusters and data centers consuming 600+ MW during peak
- Projected 18% reduction in long-term AI compute energy costs
- BofA Securities maintains Buy rating, citing energy-AI infrastructure alignment
- 2.5 GW can power over 1.5 million homes and 100,000 AI inference servers
Meta Platforms, Inc. (META) has secured 2.5 gigawatts (GW) of dedicated nuclear energy capacity through multi-year power purchase agreements with two U.S.-based nuclear operators, including a major facility in Pennsylvania and a second under development in Georgia. This energy allocation is designed to directly power Meta’s AI training clusters and data centers, which now consume over 600 megawatts of electricity during peak operations. The company plans to deploy the full capacity by 2028, with initial deliveries beginning in early 2027. The move underscores Meta’s ambition to maintain a competitive edge in artificial intelligence, where energy demands are projected to grow by over 40% annually. By locking in a stable, carbon-free energy source, Meta aims to reduce exposure to volatile grid prices and regulatory risks associated with renewable intermittency. The nuclear supply agreements are expected to lower the company’s long-term energy costs per AI compute job by approximately 18%, according to internal modeling. BofA Securities reaffirmed its Buy rating on META, citing the strategic alignment of energy security with AI scalability. The firm noted that Meta’s energy investments could serve as a model for other tech giants facing similar infrastructure constraints. Analysts estimate that the 2.5 GW of nuclear capacity is sufficient to power more than 1.5 million U.S. homes annually and would support the operation of over 100,000 AI inference servers at scale. The development has drawn attention from utility providers, environmental groups, and policymakers. While clean energy advocates applaud the reduction in fossil fuel dependence, some regulatory bodies are reviewing the implications of large-scale corporate nuclear procurement on grid reliability and regional energy distribution. The agreements also signal a growing trend of tech firms securing long-term energy contracts beyond traditional renewables.