Nvidia outlines strategies to accelerate US nuclear energy renaissance
The chipmaker argues that AI can help America build reactors faster and cheaper as decades of flat electricity demand come to an end
For decades, American electricity demand barely moved. The grid, the infrastructure behind it and much of the policy governing it were all built on the assumption that it never would.
Nvidia is now making the case in Fortune that this assumption is finished. The company is positioning itself as a key player in getting US nuclear power built at a pace the country hasn’t attempted in a long time.
The pitch is fairly simple. AI is driving up power demand, nuclear can supply steady carbon-free electricity, and AI itself might be the tool that makes building reactors less painfully slow.
What Nvidia is actually doing
The centerpiece is a project called PROMETHEUS. Nvidia and Idaho National Laboratory (INL) launched it in February 2026 under the Department of Energy’s Genesis Mission.
The goals are ambitious. PROMETHEUS aims to use AI to let reactor design, licensing, manufacturing, construction and operations happen up to twice as quickly. It also targets operational costs that are more than 50% lower.
A meaningful part of the work is computational. Nvidia is speeding up key nuclear simulation codes on its GPUs, including MOOSE and BISON.
Nvidia is also contributing AI infrastructure to support the Genesis Mission’s broader goals. Those goals are framed around making nuclear energy faster, safer and cheaper to deploy for so-called AI factories, the industry’s term for large-scale AI data centers.
The toolkit includes generative AI, digital twins and accelerated simulation. A digital twin is a detailed virtual copy of a physical system, letting operators test changes on screen before touching the real thing.
Microsoft, permits and a microreactor in Utah
In March 2026, Microsoft and Nvidia announced a joint “AI for nuclear” initiative. It targets permitting, design and operations, relying on digital twins and generative AI to speed things up.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The initiative also aims to support repeatable designs for new nuclear facilities.
One example referenced in connection with the effort showed permitting time cut by 92%.
The most tangible milestone came on July 1, 2026. Valar Atomics demonstrated its Ward 250 microreactor generating power for an Nvidia Blackwell AI chip in Utah.
It marked the first time an advanced nuclear reactor has supplied electricity to an AI system in the US. Valar and Nvidia also announced plans for a waterless data-center pilot of about 30 MW.
The policy tailwind
Nvidia’s push lines up neatly with Washington’s direction. Trump administration executive orders, including actions taken in May 2026, target quadrupling US nuclear capacity.
Nvidia and its partners have zeroed in on small modular reactors (SMRs) and microreactors. The idea is to place them alongside hyperscale AI infrastructure rather than relying solely on power shipped across the grid.
SMRs are smaller, factory-style reactors meant to be built in a more standardized way than the massive plants of previous generations. Microreactors are smaller still, closer to the scale of the Ward 250 unit that powered the Blackwell chip.
Traditional cooling methods consume significant amounts of water, and the partners are developing more water-efficient approaches, including the planned waterless pilot.
What this means
For Nvidia, the strategic logic is straightforward. Its business depends on customers building ever-larger AI data centers, and those data centers depend on electricity that the grid wasn’t designed to deliver.
The research framing these initiatives points to possible knock-on effects. If AI-and-nuclear partnerships succeed in establishing nuclear as a cornerstone of AI infrastructure, that could encourage further supportive policy, funding opportunities, grants and more favorable regulation for related technologies.
The caveats are real, though. Most of the headline numbers are targets or isolated examples.
PROMETHEUS aims for timelines up to twice as fast and costs more than 50% lower. The 92% permitting reduction comes from one referenced case. The Valar demonstration powered a single chip, and the 30 MW site is still a plan.
What to watch is whether these efforts move from pilots to repeatable deployments. Key signals include progress on the Valar waterless pilot, measurable results from PROMETHEUS at INL, and whether the Microsoft initiative’s permitting gains show up across multiple projects.