Nvidia Bets On Gates-backed Nuclear Startup To Keep Its AI Ambitions From Melting Down

Datacenter operators’ desire for cheap and clean energy to power their facilities has led to renewed interest in nuclear energy and small modular reactors (SMRs) – a tech Nvidia has just decided is worthy of investment.

The GPU giant's venture capital arm, NVentures, this week joined Bill Gates and HD Hyundai in a $650 million funding round that pumped cash into a company called TerraPower that aims to make micro-reactors a reality.

Founded in 2006 by the former Microsoft CEO, TerraPower has spent the better part of two decades developing a reactor design that's easier and cheaper to build than the massive water pressure reactors common today.

Developed in collaboration with the Department of Energy's Advanced Reactor Demonstration Project, TerraPower's Natrium plant in Wyoming will feature a 345-megawatt reactor backed by a molten salt energy storage system with a gigawatt of capacity.

Work on the non-nuclear portions of the plan began last year. But before you get too excited, it's going to be a while before TerraPower's Natrium reactor tech actually achieves fission. The company hasn't secured regulatory approval for the project and doesn't expect the Nuclear Regulatory Commission to green light the project until sometime next year.

Despite that red tape and SMRs remaining unproven, tech giants have embraced the nuclear energy. Nearly every cloud and hyperscaler has announced some kind of nuclear energy initiative. AWS is working with X-energy and Dominion Energy; Google is betting on Kairos; Microsoft is working to reignite Three-Mile Island, and Oracle has announced plans to deploy at least three SMRs with over a gigawatt of combined capacity.

TerraPower has also found a datacenter customer. In January it signed a memorandum of understanding to supply Sabey Data Centers with nuclear energy.

Nuclear energy advocates point to the fact that datacenters often use diesel or gas turbines to supplement energy they draw from the grid. For instance, xAI deployed 35 mobile gas generators to power its Colossus supercomputer in Memphis until the local utility could complete work on a pair of 150 megawatt substations.

Mobile generators and natural gas plants may provide a stopgap, but even these face health environmental challenges. This week the NAACP sued Elon Musk’s xAI over the public health risks posed by its continued reliance on gas generators.

We don't imagine folks will be thrilled to have a cluster of miniature nuclear reactors in their back yard, either, but at least they won't be filling the air with smog.

Nvidia, for its part, has a vested interest in these alternative energy projects.

"As AI continues to transform industries, nuclear energy is going to become a more vital energy source to help power these capabilities," NVentures head Mohamed Siddeek said in a canned statement. "TerraPower's nuclear reactor technologies offer innovative, carbon-free solutions to meet global energy needs while minimizing environmental impact."

In other words, the more energy datacenters can source, the more GPUs they can run.

• The launch of ChatGPT polluted the world forever, like the first atomic weapons tests
• Amazon has changed its nuclear deal in Pennsylvania to bypass grumpy regulators
• UK bets big (and small) on nuclear as datacenter demand expected to climb
• Meta just saved an Illinois nuclear plant that was set to be mothballed

TerraPower expects to get regulatory approval to move forward with its first Wyoming plant next year, but doesn’t expect its reactor will export power until 2030.

TerraPower originally planned to power up the plant in 2028 but work on the project was delayed after Russia’s invasion of Ukraine disrupted its fuel supply.

Even if TerraPower’s plant works, that doesn't mean it or anyone else's SMRs will be economically viable. NuScale, which was among the first SMR aspirants, abandoned its project after costs increased.

In fact, skeptics contend that SMRs will never take off. Last year the Institute for Energy Economics and Financial Analysis determined that, despite the hype, SMRs were "too expensive, too slow to build, and too risky to play a significant role in transitioning away from fossil fuels." ®