Nvidia Targets Nuclear Energy to Power AI Infrastructure

Nvidia is positioning itself to accelerate the expansion of US nuclear energy, arguing that AI can help build reactors faster and more affordably. The company is responding to rising electricity demand from AI data centers, which it refers to as AI factories, by integrating nuclear power as a reliable, carbon-free energy source.

The core of this effort is the PROMETHEUS project, launched in February 2026 by Nvidia and Idaho National Laboratory under the Department of Energy’s Genesis Mission. The project aims to use AI to cut reactor design, licensing, manufacturing, and construction times by up to half, while targeting a reduction in operational costs of more than 50%.

Nvidia is providing the computational infrastructure for these goals, using its GPUs to accelerate nuclear simulation codes such as MOOSE and BISON. The toolkit also utilizes generative AI and digital twins, which allow operators to test system changes in a virtual environment before physical implementation.

In March 2026, Nvidia and Microsoft announced a joint initiative to apply these technologies to nuclear permitting, design, and operations. One example cited in the initiative showed a 92% reduction in permitting time.

A milestone occurred on July 1, 2026, when a Valar Atomics Ward 250 microreactor successfully powered an Nvidia Blackwell AI chip in Utah. This marked the first instance of an advanced nuclear reactor supplying electricity to an AI system in the US. The companies are now planning a 30 MW waterless data-center pilot.

Nvidia’s strategy aligns with recent US policy, including May 2026 executive orders from the Trump administration that aim to quadruple the country's nuclear capacity. The focus is on small modular reactors and microreactors, which can be placed directly alongside AI infrastructure to bypass grid limitations.

While these initiatives aim to establish nuclear energy as a foundation for AI, the company notes that many of the figures are targets or based on isolated examples. Future progress will depend on whether these pilot projects can transition into repeatable, large-scale deployments.