NVIDIA unveils science agents and Rubin HPC GPU
NVIDIA introduced Nemotron for Science, the Rubin HPC GPU and an expanded Vera CPU platform at SC26 in Chicago. Ian Buck, vice president of hyperscale and high-performance computing, outlined tools designed to let AI agents plan experiments, execute code and verify results while keeping researchers’ data and models on their own systems.
Nemotron for Science combines the Nemotron 3.5 Super VL open model with an open chemistry dataset and post-training recipes for chemistry and life sciences. It can run on a DGX Spark, workstation or supercomputer. Buck said an Argonne National Laboratory agent used the toolkit to plan and test hypotheses, select simulations and identify new binding sites. Separately, a KTH Royal Institute of Technology agent repaired 81% of broken GROMACS test cases during an overnight run.
Rubin HPC targets science, engineering, analytics and AI, with availability expected in 2028. The German Climate Computing Center selected Vera CPUs for its forthcoming supercomputer, reflecting CPUs’ role in executing agents’ tool calls. NVIDIA also announced a quantum computing co-design collaboration with IBM and Fermilab for error-correction and applications research. Buck said the new CUDA-Q decoding algorithm TCBP-JOBS reduces error-correction research computation from 12 weeks to 45 minutes.