NVIDIA Vera CPU Opens the Way for Agentic Scientific AI at Los Alamos National Laboratory
NVIDIA Blog · Chris Porter · 2026-06-22
Los Alamos National Laboratory is deploying three new NVIDIA Vera CPU-powered supercomputers — Mission, Vision, and Veritas — expected operational in 2027, with Vera delivering 7x performance gains over x86 predecessors on agentic AI workloads and hosting LANL's URSA autonomous scientific agent framework.
Appears in
Extraction
Topics: supercomputinghpcagentic-ainvidia-veranational-labs
Claims
- LANL's three planned supercomputers — Mission, Vision, and Veritas — will combine NVIDIA Vera CPUs, Rubin GPUs, and Quantum-X800 InfiniBand in the HPE Cray GX5000 architecture, targeting 2027 operational dates.
- NVIDIA Vera CPU delivered 7x higher performance on LANL's URSA agentic AI workloads compared to CPUs in the Crossroads x86 supercomputer.
- A single Vera CPU outperforms a single-socket x86 CPU by over 3x on Monte Carlo simulation workloads while providing more than 4x the memory per core and 6x the memory per node.
- LANL's URSA framework is a modular, feedback-driven agentic AI system designed to autonomously brainstorm hypotheses, plan experiments, run simulations, and analyze results.
Key quotes
Researchers are adding a new tool for science with AI agents that can form hypotheses, choose tools, launch simulations, analyze outputs and refine the next step.