Google DeepMind Joins DOE's Genesis Mission to Bring AI to 17 National Labs

Google DeepMind will give scientists at all 17 U.S. Department of Energy National Laboratories accelerated access to its frontier AI models, starting today with AI co-scientist on Google Cloud, as part of the White House's Genesis Mission.

The Genesis Mission is a national effort to use AI to transform how scientific research is conducted across the U.S. government's laboratory system. It mobilizes the DOE's 17 National Laboratories, industry and academia to build what Google describes as an integrated discovery platform, with the stated goal of accelerating breakthroughs in energy, scientific discovery and national security. Google and the DOE say they are supporting the Administration's target of harnessing AI and advanced computing to dramatically expand the productivity of American research and innovation within a decade.

The stakes are considerable. The national labs handle research problems of unusual scale — simulating fusion plasma dynamics, searching vast candidate spaces for new materials, and processing rapidly growing volumes of scientific data and literature. Google DeepMind argues that modern deep learning methods are suited to compress the time these discoveries would otherwise require.

What the labs get, and when

The first tool available under the accelerated access program is AI co-scientist, a multi-agent virtual scientific collaborator built on Gemini and trained on Google's TPUs. The system is designed to help scientists synthesize large bodies of information and generate novel hypotheses and research proposals, with the aim of accelerating scientific and biomedical discoveries.

Google points to existing biomedical results as evidence of its potential. AI co-scientist proposed drug repurposing candidates for liver fibrosis that were validated through laboratory experiments, and predicted antimicrobial resistance mechanisms that matched experiments before those experiments were published. Google frames this as compressing hypothesis development "from years to days." The system is also showing early promise in physics, chemistry and computer science, according to the company.

In early 2026, the access program expands to three additional systems:

• AlphaEvolve — a Gemini-powered coding agent for designing advanced algorithms. Google says it has already improved the efficiency of Google's data centers, chip design and AI training processes, including training the large language models underlying AlphaEvolve itself. The company believes it could be transformative in materials science, drug discovery and energy.
• AlphaGenome — a model for understanding the non-coding portion of DNA, intended to speed up genome biology research and improve disease understanding. Google says that with more plant genome data, AlphaGenome could be extended to crop resistance, sustainable biofuels and advanced biomaterials.
• WeatherNext — a family of weather forecasting models. A partnership with the U.S. National Hurricane Center already supports cyclone forecasts and warnings.

DOE and the National Laboratories will also get access to Gemini for Government, which combines Google's accredited AI-optimized cloud with its Gemini model family, including Gemini 3 with what the company calls state-of-the-art reasoning and multimodal understanding.

A track record, and a Nobel

Google is keen to frame the partnership as a continuation of an existing relationship between its AI research and the national lab system. Brookhaven National Laboratory's foundational work on the Protein Data Bank was crucial to the development of AlphaFold, the protein structure prediction system whose creation earned DeepMind's Demis Hassabis and John Jumper a share of the 2024 Nobel Prize in Chemistry. The AlphaFold Protein Database has since been used by more than three million scientists in over 190 countries, contributing to work ranging from malaria vaccines to gene therapies.

That history matters for how the new program is judged. AlphaFold is the clearest existing case study for the claim that AI can compress scientific timelines; the Genesis partnership effectively bets that agentic tools like AI co-scientist and AlphaEvolve can repeat that pattern across physics, materials and energy research at scale.

Google characterizes the announcement as "the beginning of an enduring partnership in AI for Science that we will look to grow and expand in the months and years ahead." Beyond tool access, DeepMind says it will explore research collaborations with the National Laboratories in fusion energy, new materials discovery and earth science.

The companies and agencies involved have set a decade as the horizon for measurable gains in research productivity. Early 2026, when AlphaEvolve, AlphaGenome and WeatherNext reach the labs, will be the first concrete checkpoint for whether that timeline is realistic.