Los Alamos has begun early testing of the first NVIDIA Vera CPU server, giving researchers their first hands-on evaluation platform for the future Mission and Vision supercomputers.
Engineers will use the system to validate software, tune scientific codes and measure real workloads before full deployments of the HPE-built machines later this decade.
Mission and Vision are designed around NVIDIA’s Vera Rubin platform, pairing Vera CPUs with Rubin GPUs for high-performance computing and agentic AI research workflows.
Ben Santos said the new systems should deliver more than 3 times Crossroads’ CPU performance, over 4 times memory per core and more than 12 times the AI performance of Venado’s Hopper GPUs.
The first purpose-built Mission architecture, Starlight, is due later in 2026, with full Mission and Vision deployments scheduled for 2027 and 2028.
With agentic AI making decisions, who is accountable when the supercomputer is wrong?
As AI automates scientific discovery, what is the new role for human researchers?
LANL Unveils Mission and Vision: NVIDIA Vera Rubin Supercomputers to Deliver 8 EXAFLOPS for U.S. Science and Security
Overview
In July 2026, Los Alamos National Laboratory (LANL) received its first NVIDIA Vera CPU server, marking a foundational step for its Mission and Vision supercomputing projects. This milestone is the result of a strategic codesign collaboration between NVIDIA and Hewlett Packard Enterprise (HPE), reflecting a shared commitment to advancing high-performance computing. The Vera CPU server, as part of the Vera Rubin platform, is engineered to meet the growing demands of modern scientific computing, which now relies on a blend of simulation, artificial intelligence, and agentic reasoning. The platform delivers critical memory bandwidth, energy efficiency, and AI acceleration, empowering LANL to tackle complex scientific challenges.