The bottleneck for AI stopped being chips a while ago. It's power. Nvidia's plan to fix that now involves a small nuclear reactor in the Utah desert.
Last week, Valar Atomics announced that its Ward 250 reactor had achieved zero-power criticality, meaning it can sustain a nuclear chain reaction at essentially no power output. It's a tiny thing, rated at 100 kilowatts thermal, which is orders of magnitude below what a real data center actually needs. But that's not really the point. The point is what Valar built the reactor for, and who they built it with.
"Nine months ago, this was an empty site. Today, there's a critical reactor on it," Valar founder and CEO Isaiah Taylor said. Valar is one of the startups moving through the Department of Energy's Reactor Pilot Program, and it just partnered with Nvidia on what both companies are calling the future of AI infrastructure: dedicated, behind-the-meter nuclear power sitting directly next to a data center, with no grid in between.
The nuclear announcement lands the same week Nvidia released the Vera Rubin DSX reference design, a full architectural playbook for what Nvidia calls "AI factories." The headline feature is cooling. Rubin systems run on a closed liquid loop with coolant entering the chip at 45 degrees Celsius, which is warm enough that a lot of the world's data centers don't need chillers or cooling towers at all. Just radiators outside.
Which brings us back to why the Valar partnership matters. Nvidia's cooling numbers are dramatic on paper:
- Zero water consumption, versus roughly 2.6 million gallons per megawatt per year for a traditional cooling-tower setup.
- Around $4 million a year in savings on cooling for a 50-megawatt facility.
- Three times the rack density, because you can rip out the fans that used to make the room sound like a jet engine.
"Once the watts per chip crossed a certain level, liquid cooling became mandatory," said Richard Whitmore, president of Motivair, the Schneider Electric division building much of this hardware. Modern Nvidia chips run hot enough that air just can't keep up.
Stack all of that together and the pitch to hyperscalers starts to make sense. You get a data center that doesn't need to be near water, doesn't need to be near a substation, and doesn't need to fight local governments over grid capacity. It just needs a reactor next door. John Josephakis, an Nvidia global VP, framed the Valar deal as exploring how "behind-the-meter, waterless advanced nuclear systems could support future AI factories."
Nvidia isn't alone in this. Constellation is restarting Three Mile Island for Microsoft. Amazon and Google both have nuclear PPAs signed. What's different about the Valar setup is the scale and the ownership model. Instead of buying grid power from a utility that happens to have a reactor, hyperscalers would own the reactor as part of the facility, or at least sit on the other side of the same fence.
The catch is that nothing about this is close to production ready. Ward 250 puts out 100 kilowatts. A single Rubin rack draws more than that. Small modular reactors also come with the highest electricity cost of any behind-the-meter option modeled in recent MDPI research, running around 25% above grid baseline. Advanced nuclear is expensive, slow to license, and has a fuel supply chain the DOE itself describes as still "under construction." A demonstration reactor in Utah is not a commercial reality in Virginia.
Nvidia isn't really selling nuclear reactors here. It's selling a story to hyperscalers about how the current build-out actually works at scale. Chips are the easy part now. The hard part is where you get 10 gigawatts of clean power without waiting a decade for grid interconnects. If Valar and the other advanced reactor startups can turn a critical test reactor into something that actually powers a data center by 2030, the map of where AI gets built starts to look very different, with the desert winning and the coasts losing. If they can't, the industry has a very expensive problem it has been quietly pretending is already solved.
