DOE Opens Free Supercomputer Access for AI Research Teams
March 25, 2026 · 2 min read
Arthur Griffin
The Department of Energy's National Energy Research Scientific Computing Center (NERSC) is accepting proposals for its 2026 AI for Science program, offering researchers free access to one of the world's most powerful supercomputers to advance artificial intelligence applications in scientific discovery.
Submissions received by April 30, 2026 will receive full consideration, though proposals are reviewed on a rolling basis until computing resources are exhausted.
What Perlmutter Offers Accepted Teams
Selected projects receive up to 10,000 GPU node hours on the Perlmutter system — each node containing four NVIDIA A100 GPUs — along with 20,000 CPU node hours for generating AI-ready datasets. NERSC also provides dedicated file system storage and staff consulting on system optimization.
The awards cover the full NERSC 2026 Allocation Year through January 19, 2027, and teams that demonstrate effective use can request additional allocations.
Who Should Apply
The call is deliberately broad. Prior NERSC experience is not required, making this a rare on-ramp for teams at universities, national labs, and independent research organizations that lack their own GPU infrastructure. NERSC is looking for teams with "expertise in deep learning for science, a deep understanding of the scientific domain, and demonstrated proof-of-concept results."
Proposals must outline a well-defined scientific objective, computational resource requirements, and specific deliverables with timelines. Projects are evaluated on scientific significance, relevance to DOE missions, technical feasibility, and the team's track record.
A Growing Pipeline for AI-Driven Science
This call sits within a broader DOE strategy that now includes $68 million in dedicated AI research grants and the $320 million Genesis Mission, creating a pipeline where researchers can pair federal grant funding with world-class computing infrastructure.
For teams developing AI foundation models, climate simulations, drug discovery tools, or materials science applications, the barrier to entry has never been lower. The application form is a single submission — no multi-stage review process.
Grant seekers tracking AI-for-science funding can find deeper analysis of DOE's computing strategy on the Granted blog. Questions can be directed to nersc-ai@lbl.gov.