At the SC26 conference in Chicago on November 17, 2026, NVIDIA unveiled a collection of advanced tools and hardware designed to usher in what the company calls the ‘Age of Agents’ in scientific discovery. Ian Buck, NVIDIA’s Vice President of Hyperscale and High-Performance Computing, presented a keynote announcing a portfolio of technologies meant to accelerate and automate complex research and simulation processes with the help of artificial intelligence.

Nemotron for Science Empowers Research Autonomy

Chief among the new launches is Nemotron for Science, a toolkit combining the open model Nemotron 3.5 Super VL, an extensive open-access chemistry dataset, and guidance for post-training specialization. This tool allows researchers to develop their own domain-specific AI agents with the capacity to plan experiments, run simulations, and interpret results entirely on their own computing infrastructure—whether that be on NVIDIA’s DGX Spark systems, high-powered workstations, or personal supercomputers. In demonstrations, such agents showed the ability to autonomously design and run experimental workflows, identifying new chemical binding sites based on a single scientist’s prompt.

Hardware Expansion: Rubin HPC GPU and Vera CPU

NVIDIA also introduced the Rubin HPC GPU, a new processor in the Vera Rubin lineup, tailored specifically for scientific computing, high-precision analytics, and AI-heavy workloads. The Rubin GPU extends support across a range of computing precisions (FP64 to FP4), adds hardware for sparse-data acceleration, and leverages seventh-generation NVLink for high interconnect speeds. The chip boasts over 50,000 CUDA cores—triple that of its predecessor, Hopper—and is slated for public availability in 2028.

Meanwhile, the Vera CPU platform is expanding into new research domains. The German Climate Computing Center (DKRZ) confirmed that it will power its HLRE-5 supercomputer with Vera CPUs, aiming for deployment in early 2028—a move that demonstrates growing industry adoption of NVIDIA’s high-performance central processors.

Quantum Computing Partnerships

Charting a course toward quantum-era research, NVIDIA announced a collaboration with IBM and Fermi National Accelerator Laboratory. The joint initiative aims to create a quantum computing co-design platform supporting both error-correction technology and scientific application development. Plans involve pairing NVIDIA DGX B200 systems with IBM quantum computing hardware, including machines powered by IBM’s Quantum Nighthawk, a move that underscores the growing overlap between classical and quantum computing realms in cutting-edge research.

Significance and Next Steps

NVIDIA’s announcements build on two decades of leadership in the field, from the launch of CUDA in 2006 to the development of over a thousand CUDA-X libraries. The company’s push to make AI agents a core part of scientific research infrastructure has both immediate and long-term implications for fields such as climate science, chemistry, and engineering.

While many of these products—including Rubin HPC GPU and the expanded Vera CPU platform—are not scheduled for release until 2028, and integrations with quantum computing partners remain in early phases, NVIDIA’s announcement sets a clear agenda for the next era of computational science. Details about practical deployment and in-lab integration of these technologies will be closely watched as scientists look for ways to harness the full promise of agent-led discovery.