TerraPower's energy-storage edge could give it an advantage powering AI data centers

Nuclear power startups have been pitching themselves as the antidote to what ails AI data centers: power that’s always available. Bill Gates-founded TerraPower is one of the latest to throw its hat into that ring, with Bloomberg reporting that the startup plans to announce its first data center project this year.

TerraPower did not say who the customer will be, though in January it announced that Meta had agreed to buy eight of its Natrium power plants. The data center project, expected to break ground in 2027, would be the company’s second power plant, with its first already under construction in Wyoming.

Not every nuclear reactor is suited to data center duty, but TerraPower possesses one key advantage — energy storage — that promises to give it an edge over competitors, thanks to renewable power sources like wind and solar. Nuclear reactors, TerraPower’s included, work best running at full tilt: of all power plant types, nuclear reactors have the highest capacity factor, generating at maximum power 92.5% of the time in the US. But existing reactors are slow to ramp up and down, capable of increasing or decreasing only about 5% of their total rated output per minute, according to the National Laboratory of the Rockies (NLR). New small modular reactors (SMRs) can react faster, about 10% of their rated output per minute, per NLR — but running at reduced capacity isn’t ideal, since it’s hard to make money when not generating electrons.

That’s true of any power plant, but especially nuclear, which has the highest capital expenditures of any generating technology. Startups are hoping mass manufacturing of SMRs will bring capex down, though that has yet to be proven, and could take a decade or more to pay off — so it makes sense to run early plants at peak capacity as often as possible. For data centers, especially those relying on behind-the-meter power, that poses a challenge: their loads, particularly when training AI or responding to prompts, can sink and soar quickly as GPUs respond to tasks. The swings are demanding enough that natural gas turbines have been breaking under the stress, pushing operators toward large, costly battery banks to smooth the curve.

TerraPower designed its 345-megawatt molten salt-cooled reactor to work around those challenges. A key consideration was ensuring the reactor could complement intermittent sources like wind and solar — meaning it needed to ramp up and down quickly. While TerraPower had renewable power, not data centers, in mind when it sketched the design, the two turned out to be similarly intermittent, just on different sides of the equation. To ramp quickly, TerraPower doesn’t change the reactor’s power output; instead it keeps splitting atoms continuously, storing the extra heat in a giant vat of molten sodium. When power demand spikes, the plant taps that reservoir to generate more steam and spin the turbines, so expensive equipment keeps working even when demand is low, letting the company amortize its investment over more operational hours.

Why it matters

By pairing nuclear’s high capacity factor with an energy-storage buffer built for renewables, TerraPower is positioning its reactor design as compatible with the volatile power draw of AI training — a niche that could give it an edge as nuclear startups race to sign data center customers.