Nuclear power startups have been positioning themselves as the answer to one of AI data centers' biggest challenges: the need for a constant, reliable power supply. TerraPower, the company founded by Bill Gates, is the latest to enter this arena. According to a Bloomberg report, the startup plans to announce its first data center project this year. While TerraPower hasn't disclosed who the customer will be, it was revealed in January that Meta had agreed to purchase eight of its Natrium power plants. The upcoming data center project, expected to break ground in 2027, will mark TerraPower's second power plant, following its first, which is already under construction in Wyoming.

Not every nuclear reactor is a good fit for data center operations, but TerraPower has a distinct advantage that could set it apart from the competition: energy storage. Interestingly, this edge comes thanks to renewable energy sources like wind and solar. Nuclear reactors, including TerraPower's, perform at their best when running at full capacity. In the U.S., nuclear reactors boast the highest capacity factor of any power generation type, operating at maximum output about 92.5% of the time. However, this is partly because they need to—traditional reactors are slow to adjust their output, changing by only about 5% of their total rated capacity per minute, according to the National Laboratory of the Rockies. Newer small modular reactors (SMRs), which many startups are focused on, can respond more quickly, adjusting at roughly 10% of their rated output per minute.

Still, operating at reduced capacity is far from ideal. It's hard to turn a profit when you're not producing electricity, and this is especially true for nuclear power, which carries the highest capital expenditures of any generating technology. Startups hope that mass manufacturing SMRs will eventually lower these upfront costs, but that hasn't been proven yet. Even if it works, it could take a decade or more to see the benefits. Every startup acknowledges that its early power plants will be pricey, so it makes sense to run them at peak output as often as possible.

For data centers, particularly those relying on behind-the-meter power, that's a real challenge. Their energy demands can spike and plummet rapidly, especially when training AI models or responding to prompts, as GPUs react to the workload. These swings are so intense that natural gas turbines have been known to fail under the strain. To smooth out these fluctuations, data centers often need large battery banks, which add even more costs.

TerraPower designed its 345-megawatt molten salt-cooled reactor with these issues in mind. A key priority was ensuring the reactor could work alongside intermittent energy sources like wind and solar, meaning it needed to ramp up and down quickly. While TerraPower had renewable power in mind—not data centers—when it drew up its plans, the two are actually quite similar in their intermittent nature, just on opposite sides of the equation.

To achieve rapid ramping, TerraPower doesn't simply increase or decrease the reactor's power output. Instead, it keeps splitting atoms continuously, and the excess heat is stored in a massive vat of molten sodium. When power demand surges, the plant can draw on that reservoir to generate extra steam and spin the turbines. This way, the expensive equipment keeps running even when demand is low, allowing the company to spread its investment over more operational hours. This approach takes the best of nuclear power—its high capacity factor—and pairs it with energy storage technology that lets TerraPower operate smoothly on a renewable-heavy grid or when connected to a data center. It's a flexible strategy that could give the startup a significant edge in the race to power AI.