Los Alamos Moves 11,000 Supercomputer Nodes to OpenCHAMI
Los Alamos National Laboratory has moved most of its production supercomputing onto OpenCHAMI, an open-source framework for building and configuring high-performance computing systems. Seven production clusters totaling more than 11,000 compute nodes now run on the system, the laboratory announced on November 13, 2026. OpenCHAMI began in 2023 as an open collaboration among Los Alamos, NERSC, CSCS, HPE and the University of Bristol. The announcement doesn't quantify how much the move cuts downtime or operating cost.
The migration is the news, not a benchmark. Los Alamos didn't claim the machines got faster; it changed who controls the software that builds and runs them, and it chose a framework anyone can read.
Los Alamos National Laboratory has finished moving most of its production supercomputing onto open-source software it helped build. Seven of the laboratory's production clusters, totaling more than 11,000 compute nodes, now run on OpenCHAMI, a cloud-native framework for managing high-performance computing systems. The laboratory announced the completed move on November 13, 2026.
OpenCHAMI's job is the unglamorous layer underneath the science: building clusters, configuring them, and managing operations and change across thousands of machines that have to behave as one computer. When a cluster of that size needs an upgrade or a repair, the framework is what decides whether the work takes an afternoon or a shutdown. Los Alamos now runs that layer on software whose code anyone can inspect.
The framework wasn't built by a vendor. OpenCHAMI launched in 2023 as an open community collaboration whose founding members include Los Alamos, the National Energy Research Scientific Computing Center (NERSC), the Swiss national supercomputing center CSCS, HPE and the University of Bristol. A national laboratory putting its production fleet on a community codebase is a bet that shared maintenance beats a proprietary contract.
The announcement is careful about what it doesn't claim. It doesn't quantify a reduction in downtime for upgrades, and it doesn't publish a performance comparison against the environment the clusters ran on before. The machines didn't get new processors this week. What changed is the management layer, and the laboratory is presenting the migration itself — completed, at production scale — as the result.
Scale is the credential here. Eleven thousand nodes across seven systems is not a pilot. It's the working fleet of one of the oldest supercomputing sites in the world, the laboratory that has run first-of-their-kind machines since the 1940s. A framework that survives that fleet's daily operations has passed a test no benchmark reproduces.
Other laboratories are the obvious next audience. NERSC and CSCS were founding members, which means the two other national-scale centers in the collaboration have their own reasons to adopt what Los Alamos just proved in production. Open-source infrastructure spreads exactly this way: one big site goes first, publishes the scars, and the rest of the field stops having to guess.
The conversation.
Loading comments…

A little more perspective.
Reading is free. Verified members can comment and reply for $2 a month.