NERSC issues 2026 call for AI for Science proposals on Perlmutter

The National Energy Research Scientific Computing Center has opened a call for proposals to use its Perlmutter supercomputer for artificial intelligence projects in science. The center seeks teams with deep learning expertise, domain knowledge and prior proof of concept results. The call is open to anyone and does not require prior NERSC usage. Accepted projects will receive up to 10,000 Perlmutter GPU node hours and associated storage quotas. Each GPU node contains four A100 units. Up to 20,000 CPU node hours per project may also be awarded to generate publicly available AI ready datasets. All awards apply to the 2026 allocation year that ends January 19, 2027. NERSC staff will assist with effective resource use but not model development. Proposals must define project scope with scientific background, objectives, methodology and community impact. They must also specify resource needs, team expertise and a clear timeline with deliverables. Projects will be evaluated on scientific significance and innovation, relevance to NERSC, Lawrence Berkeley National Laboratory and Department of Energy Office of Science missions, technical feasibility and readiness to use large scale resources at full capacity. Proposals are reviewed on a rolling basis until resources are exhausted. Submissions received by April 30, 2026 will receive full consideration. NERSC serves as a Department of Energy Office of Science user facility at Lawrence Berkeley National Laboratory and supports research across multiple scientific fields.