Meta TerraPower nuclear deal secures 4 GW for AI data centers
Meta has entered an agreement with TerraPower to utilize eight Natrium nuclear reactor units to provide up to 4 GW of power for its AI data center infrastructure. The reactor design features an integrated molten salt energy storage system, allowing for flexible power output to meet the high and variable energy demands of AI workloads.
Key Takeaways
- Meta targets 6.6 GW of total nuclear capacity to fuel its long-term AI infrastructure requirements
- Natrium reactors feature thermal batteries that increase power output by 45% without additional fuel consumption
- Construction of the first utility-scale advanced plant, Kemmerer Unit 1, began in Wyoming following April 2026 NRC approval
- TerraPower signed a memorandum of understanding with Sabey Data Centers for potential deployments in Texas and the Rockies
Why It Matters
The adoption of Natrium reactors marks a shift toward dispatchable carbon-free energy that can match the volatile power cycles of AI training clusters. By integrating molten salt storage, Meta gains a strategic buffer against grid instability that traditional baseload nuclear cannot provide. This move intensifies the energy arms race among hyperscalers, following similar nuclear investments by Google with Kairos Power and Microsoft at Three Mile Island. As streaming and AI workloads converge on shared infrastructure, securing dedicated, scalable power becomes a primary competitive moat. Watch for the 2032 delivery timeline to see if regulatory hurdles or AI infrastructure spending constraints delay these first-of-a-kind reactor deployments.
Additional Context
TerraPower's Natrium reactor is attracting multiple hyperscaler commitments as AI data center power demand outpaces conventional grid capacity. In June 2025, TerraPower broke ground on its first Natrium reactor at a retired coal plant site in Kemmerer, Wyoming, marking the first new commercial nuclear construction in the U.S. outside of federal national laboratories in decades. The Natrium design pairs a 345 MW sodium-cooled fast reactor with a molten salt thermal storage system that can boost output to 500 MW for up to five hours, a feature that directly addresses the bursty power profiles of AI training workloads. TerraPower has indicated it aims to have the Kemmerer unit operational by 2030, with subsequent units following in the early 2030s. The competitive landscape for nuclear-powered data centers has intensified sharply over the past year. In October 2024, Google signed the world's first corporate agreement to buy power from multiple small modular reactors with Kairos Power, targeting 500 MW from six to seven reactors with the first unit online by 2030 and additional deployments through 2035. Microsoft took a different approach in September 2024, announcing a 20-year agreement with Constellation Energy to restart Three Mile Island Unit 1, which will supply 835 MW of carbon-free power to Microsoft data centers. Amazon, meanwhile, acquired the Susquehanna nuclear-powered data center campus in Pennsylvania for $650 million in March 2024, securing direct access to 960 MW of existing nuclear generation. These moves collectively signal that hyperscalers view firm carbon-free power as a strategic asset rather than a procurement line item. The technical and regulatory case for Natrium's molten salt storage system centers on load-following capability that conventional light-water reactors cannot match. Kairos Power, which also uses molten salt cooling rather than water, began construction on its Hermes Low-Power Demonstration Reactor in Oak Ridge, Tennessee in July 2024, giving it a head start on regulatory engagement with the NRC. However, Reuters reported that regulators' busy workload and potential uranium supply obstacles may delay nuclear permitting for data center deals, and local opposition remains a factor for new reactor siting. For Meta's eight-unit Natrium deployment, the 2032 delivery timeline will depend on whether the NRC can process construction permits at a pace that matches AI data center capacity demand curves, especially as AI infrastructure power constraints continue to throttle cloud growth.
Read full article at cryptobriefing.com
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source