TerraPower's 345-megawatt molten salt-cooled reactor stores excess heat in molten sodium, allowing it to adjust power output to match the wild demand swings of AI data centers without relying solely on expensive battery systems. The Bill Gates-founded startup plans to announce its first data center project this year, representing a shift in how nuclear power can compete for AI infrastructure contracts.

Why Do AI Data Centers Struggle With Traditional Nuclear Power?

AI data centers face an energy challenge that traditional power plants struggle to solve. When large language models (LLMs) train or respond to user prompts, their power consumption swings dramatically as graphics processing units (GPUs) spin up and down to handle tasks. These demand spikes are so severe that natural gas turbines have actually broken under the stress.

Nuclear reactors are engineered to run at full capacity constantly. In the United States, nuclear plants operate at a 92.5% capacity factor, meaning they generate maximum power nearly all the time. But this strength becomes a weakness when paired with data centers. Nuclear reactors are slow to adjust output, capable of increasing or decreasing only about 5% of their total power per minute, according to the National Laboratory of the Rockies. Even newer small modular reactors (SMRs), which many startups are pursuing, can only ramp at about 10% per minute.

To smooth out the mismatch between steady nuclear output and fluctuating data center demand, facilities typically rely on massive battery banks, which significantly increase project costs. For nuclear power to make economic sense, it needs to run at peak capacity as often as possible, since nuclear has the highest capital expenditures of any generating technology.

How Does TerraPower's Thermal Storage System Address This Problem?

TerraPower designed its reactor with a thermal storage solution that sidesteps the traditional ramp-up challenge. Rather than increasing or decreasing the reactor's power output, the plant keeps splitting atoms at full speed. The extra heat generated during low-demand periods gets stored in a massive reservoir of molten sodium. When power demand spikes, the plant taps that thermal storage to generate additional steam, spinning the turbines faster without slowing the nuclear reaction.

This approach was originally conceived to help nuclear plants work alongside intermittent renewable energy sources like wind and solar. The two systems face similar challenges: renewables fluctuate based on weather, while data center loads fluctuate based on computing tasks. By keeping the reactor running at full capacity while using thermal storage to absorb and release energy, TerraPower achieves the best of both worlds.

What Advantages Does This Design Offer?

• Continuous Equipment Operation: The expensive reactor equipment keeps working even when demand is low, allowing TerraPower to spread its massive upfront investment across more operational hours and improve profitability.
• Reduced Battery Requirements: Data centers can reduce their reliance on large battery banks to smooth power fluctuations, lowering total project costs and operational complexity.
• Flexible Deployment Options: The system works equally well on renewable-heavy grids or when directly connected to a data center, making it adaptable to different scenarios.
• Aligned with Data Center Needs: The thermal storage approach mirrors the intermittent nature of data center loads, creating a natural fit between supply and demand patterns.

TerraPower's first reactor is already under construction in Wyoming, and the company announced in January that Meta had agreed to purchase eight of its Natrium power plants. The data center project expected to break ground in 2027 would be TerraPower's second power plant deployment, signaling confidence in the technology's viability.

How to Understand TerraPower's Competitive Position

• Cost Structure Challenge: Every nuclear startup acknowledges that early power plants will be expensive, with hopes that mass manufacturing of small modular reactors will eventually bring costs down, though that reduction could take a decade or more to materialize.
• Operational Advantage: While TerraPower's molten salt approach doesn't solve the cost problem directly, it does solve the operational problem that makes nuclear uneconomical for data centers by allowing the reactor to run at peak efficiency while meeting variable demand.
• Market Timing: The startup's ability to make the economics work without waiting for manufacturing breakthroughs could give it an advantage over competitors in the intensifying competition to power artificial intelligence infrastructure.

The nuclear power industry has long struggled with the economics of early-stage deployments. TerraPower's thermal storage innovation represents a practical solution to one of the most pressing challenges facing nuclear-powered data centers: how to keep expensive reactors running at full capacity while serving loads that fluctuate unpredictably.