BREAKING NEWS

TerraPower Uses Molten Salt Tech for AI Data Center Power

Bill Gates-founded nuclear startup TerraPower plans to announce its first data center project this year, leveraging a 345-megawatt reactor equipped with molten salt energy storage. As reported by TechCrunch, the design helps manage rapidly fluctuating AI energy demands without reducing overall reactor output.

QuickTool Team

✓

Aug 19, 2026•3 min read•Source: TechCrunchAI-assisted summary · Automatically reviewed by the QuickTool Quality Pipeline

TerraPower Uses Molten Salt Tech for AI Data Center Power

⚡ In Short

Meta agreed to buy eight Natrium power plants from TerraPower in January.
The 345-megawatt molten salt reactor stores surplus thermal energy during low power demand.
TerraPower's second nuclear plant project is scheduled to break ground in 2027.

What Happened?

According to Bloomberg and TechCrunch, TerraPower intends to break ground on a second power plant facility in 2027 dedicated to data centers, following an earlier project in Wyoming. Meta previously agreed to purchase eight Natrium reactors from the startup. TerraPower's 345-megawatt molten salt-cooled reactor continues splitting atoms continuously, storing unneeded thermal energy in a reservoir of molten sodium. When electrical demand spikes, the system taps this stored heat to generate steam for additional power generation.

Key Highlights

Meta agreed to buy eight Natrium power plants from TerraPower in January.

The 345-megawatt molten salt reactor stores surplus thermal energy during low power demand.

TerraPower's second nuclear plant project is scheduled to break ground in 2027.

Why It Matters

Artificial intelligence workloads vary wildly during GPU processing, straining electrical infrastructure and breaking conventional gas turbines. While nuclear power offers a 92.5% capacity factor in the U.S., reactors are costly and slow to adjust output safely. TerraPower’s thermal energy storage allows the plant to run at full capacity constantly, stabilizing power delivery to energy-heavy computing setups without relying on expensive extra battery banks.

💡 Related AI Tools

High-performance AI platforms depend heavily on stable compute environments. Infrastructure improvements that stabilize data center power supply ensure low-latency API access and uninterrupted processing for complex machine learning tasks, reducing overall operational costs across cloud-hosted tools.

Conclusion

TerraPower’s hybrid approach combines high-capacity nuclear generation with flexible thermal storage. By addressing the physical power constraints of data centers, the startup offers an adaptable energy solution designed to support the expansion of energy-intensive computing systems.

Found this news helpful? Share it with your network!