# NVIDIA Unveils AI Agents, New GPUs, and CPU Platforms at SC26 Source: https://undeniableprogress.com/up-news-e3db58768c40e46d8013/ Date: 2026-11-17 At the SC26 conference in Chicago, NVIDIA’s Ian Buck opened the company’s week with a focus on AI agents, new hardware, and expanded CPU platforms designed to accelerate scientific research. Buck, vice president of hyperscale and high-performance computing at NVIDIA, emphasized that simulation remains the bedrock of scientific computing, now augmented by AI, quantum computing, and autonomous agents. Original editorial illustration; not a product screenshot. NVIDIA introduced Nemotron for Science, a new open model combining the Nemotron 3.5 Super VL architecture with a chemistry dataset and post-training recipes for life sciences and chemistry. Researchers can run it locally on DGX Spark workstations or their own supercomputers. The details At Argonne National Laboratory, an agent built using the toolkit took a single prompt, planned and tested hypotheses, selected simulations, and identified two new binding sites. Buck also highlighted existing agent deployments: GROMACS agents from KTH Royal Institute of Technology repaired 81% of broken setups in one overnight run on a DGX Spark; SETI Institute agents plotted methanol lines and adjusted for Doppler shifts autonomously; and NVIDIA’s Nsight AI Porting Agent sped up a Weather Research and Forecasting (WRF) routine by more than five times using 150 million tokens. To support the agent loop, Buck noted the need for CPUs to handle tool calls between model steps. NVIDIA’s expanded Vera CPU platform, including the Vera CPUs selected by DKRZ for its HLRE-5 supercomputer (expected in early 2028), is designed to meet this need. The platform is already being adopted by the Texas Advanced Computing Center. A new GPU, the Rubin HPC, is also in development. Built for science, engineering, analytics, and AI, it supports FP64 to FP4 precision, sparse-data acceleration, and seventh-generation NVLink. With over 50,000 CUDA cores — about three times more than Hopper — it is expected in 2028. Buck also announced a new collaboration in quantum computing with IBM and Fermi National Accelerator Laboratory, focusing on error correction techniques. A simulation on Jülich’s JUPITER supercomputer demonstrated that magic state cultivation could reduce operations needed for quantum error correction. NVIDIA’s tools and platforms aim to empower researchers to own their data and models, Buck said, while agents act as new scientific collaborators — not just tools, but active participants in the research process. Sources Ian Buck Kicks Off NVIDIA’s Week at SC26 as Science Enters the Age of Agents Nvidia at SC26: Agents are the new scientists, and they need a lab