Dell and the US Department of Energy Are Building National AI Infrastructure for Science and Government Work

Dell Technologies and the US Department of Energy have agreed to build national AI infrastructure together, which will run scientific and government projects. The focus is not individual models but the whole chain: compute, data, supercomputers and a platform for accelerating research.

Betting on infrastructure

Michael Dell and US Under Secretary of Energy for Science Darío Gil discussed the partnership publicly. Their thesis is simple: the US already has strong labs, universities and contractors, but the next leap requires a shared compute layer that connects AI, high-performance computing and scientific instruments. In this scheme, Dell acts as the industrial builder of the infrastructure, while the DOE is the customer and coordinator of scientific tasks, from energy to national security.

This is part of the broader Genesis Mission program, which the US presents as a national platform for accelerating scientific discovery. Its goal is to link supercomputers, experimental facilities, AI systems and unique datasets into a single environment and to double the effectiveness of American R&D spending within a decade. According to Gil, the infrastructure should serve not only fundamental science but also very practical areas: power grids, new materials, biomolecular design and quantum technologies.

What it is made of

The most concrete element of the partnership is the Doudna supercomputer for the NERSC center at Lawrence Berkeley Lab. It is planned to enter service by the end of 2026, and the US Department of Energy says outright that the system may become a template for other agencies. Dell is building the machine with NVIDIA, and it is meant to deliver more than a tenfold performance gain over the current flagship, Perlmutter, so it can handle classic HPC workloads alongside AI training and inference.

• Doudna for energy, materials science and biomolecular research
• A CPU-GPU pairing of the NVIDIA Vera Rubin platform and liquid-cooled Dell servers
• A single environment for simulations, scientific data and AI models
• Use as a foundation for Genesis Mission and future government programs

In parallel, the DOE has already broken the initiative down into 26 science and technology challenges. They include faster energy grid planning, digitizing decades of nuclear data, adaptive accelerators for science, and AI-driven materials design that cuts development cycles from years to months.

In other words, the government is trying to build not just "another supercomputer" but a pipeline in which computing power is tied directly to specific applications, national priorities and measurable results.

Politics and limits

A separate political layer of this story is the dispute over who controls the rules for AI use once the technology moves into the public sector. Asked about the conflict between Anthropic and the Pentagon, Michael Dell took an uncompromising stance: if a company works with the state, it cannot fully dictate how the authorities use its tools. It is an important signal not only for defense contracts but for any large government AI procurement in the US.

"A company cannot dictate to a sovereign state how to use its tools."

The context is sensitive. The conflict between Anthropic and the Pentagon arose over restrictions on using AI for mass surveillance and fully autonomous weapons. Against that backdrop, Dell's words show that part of the infrastructure market sees the problem differently from the developers of foundation models: for the former, the priorities are scale and access; for the latter, control over usage scenarios. The deeper AI moves into science, energy and defense, the more often such contradictions will move from back rooms into public policy.

What it means

For the market, this signals a phase change: in the US, AI is moving more and more from chatbots and pilots to national computing systems built for specific scientific and government tasks. If Dell and the DOE bring Doudna and Genesis Mission to working scale, it will be more than another hardware purchase: it will be a model for how states build sovereign AI infrastructure around science, energy and security.