Exhibitor login
Nvidia 17 November 2026

NVIDIA's Advance in AI Agents at SC26

NVIDIA's Advance in AI Agents at SC26

NVIDIA's Ian Buck opened the company's week at SC26 in Chicago with a presentation on new tools for AI agents. This software can schedule experiments, execute code, and check results on researchers' systems. During his keynote address, Buck announced the launch of Nemotron for Science, the Rubin HPC GPU, and a comprehensive Vera CPU platform for high-performance computing. He also spoke about a new collaboration in quantum computing with IBM and the Fermi National Accelerator Laboratory.

Buck emphasized that simulation will remain the foundation for scientific computing, supplemented by AI, quantum technology, and now also agents. He introduced Nemotron for Science, which utilizes the open model Nemotron 3.5 Super VL and a new open chemistry database. Researchers can run this technology locally on their own systems, allowing them to own data and models. An example from Argonne National Laboratory showed that an agent was able to plan hypotheses and identify two new binding sites.

Besides Nemotron, Buck mentioned several other successful applications of agents in scientific research. For instance, an agent from KTH Royal Institute of Technology fixed 81% of the test cases for GROMACS in one night. Similarly, an agent at the SETI Institute took a simple prompt and generated complex data processing to map Doppler shifts. Moreover, an agent from NVIDIA accelerated the porting of a routine from the Weather Research and Forecasting (WRF) model to GPUs by more than 5x.

Buck explained that agents rely on CPUs for supporting calculations and that DKRZ, the German Climate Computing Center, has chosen the Vera CPUs for its new HLRE-5 supercomputer. He also announced the Rubin HPC GPU, which will be available in 2028, designed for science and engineering, featuring a significant increase in processing power. Finally, Buck discussed progress in quantum error correction and the collaboration between NVIDIA, IBM, and Fermilab for a co-design platform in quantum computing.

Read the full article from Nvidia.