National Compute plans to donate $100 million in computing credits to the Trump administration for its Genesis mission, giving government researchers access to advanced computing resources through a network backed by technology companies, universities and investors. The company expects to announce the donation Thursday at an event with White House Office of Science and Technology Policy Director Michael Kratsios.
The credits would support a government-wide effort to use artificial intelligence to accelerate scientific discovery. Two people familiar with the matter, who weren’t authorized to discuss it, said the donation would support the mission. The White House declined to comment on the donation or the computing network.
Who controls access
National Compute is building what it calls the National Compute Grid, a shared network designed to pool computing capacity from multiple providers and make it available to researchers, companies and government agencies. A draft paper reviewed by POLITICO compares the proposed grid to the power grid or interstate highway system: a way to connect resources that now sit isolated from one another.
The network would combine capacity from labs and cloud providers using Nvidia and AMD processors, as well as Google’s tensor processing units. Users could buy computing as needed, reserve capacity for larger projects or connect their own systems. They’d pay only for the power they use, according to the paper.
The company has pooled computing power from major technology companies. Chipmakers, cloud and infrastructure providers including Vultr and Crusoe, and universities back it. Venture capitalists and institutional investors, including public investors such as universities and pension funds, are financing National Compute. The company aims to pool 2 gigawatts of computing capacity by 2030; it has allocated 750 megawatts to date.
Government access and model testing
The administration’s research push comes as it seeks to expand access to computing power under what the article describes as a largely light-touch approach to AI regulation. That approach prioritizes winning a race with China to build the most advanced models. The administration calls its research effort a “golden age of science,” and officials will celebrate it at Thursday’s event.
National Compute is also launching Grid Defense, a program that lets government agencies test AI models through Marshall, an agent embedded in the grid. Agencies can set their own testing procedures before deciding whether models are safe enough to deploy, National Compute Cofounder Anjney Midha said. The company designed details to remain confidential, partly to stop frontier models from learning how to evade evaluations.
Marshall would coordinate testing and return anonymized results to researchers. The government could see the results, while labs could decide voluntarily whether to release a model. “This is continuous frontier alignment monitoring at scale,” Midha said. The approach would “enable your public stakeholders to monitor you as they need,” he said.
Private infrastructure, public mission
The program would likely depend on buy-in from AI labs and the federal government. National Compute said it created the program in response to a series of high-profile incidents involving AI agents, including attacks on companies that AI labs reported over the summer.
Grid Defense entered invite-only early access in June with some government agencies, according to the paper. Starting this week, the company plans to let public-sector users with a .gov or .mil email address, along with students and faculty with .edu addresses, request access through a scheduling system. Each public-sector user will receive $100 in National Compute credits, separate from the $100 million donation for the Genesis mission.
“We think there should be thousands more Anthropics and OpenAI's,” Midha said. “But to unleash that kind of innovation, which is extraordinarily profitable for the public, we need to reduce the barrier to entry, increase compute access, and allow the researchers and scientists to do what they do best.”