
According to NVIDIA's announcement at Supercomputing 2026 (SC26), the company has introduced its new Rubin HPC GPU, designed to accelerate scientific research, engineering applications, and advanced data analytics. Expected to launch in 2028, the new GPU highlights NVIDIA's strategy of integrating artificial intelligence (AI) with traditional high-performance computing (HPC) to advance scientific discovery and industrial innovation.
As major technology companies continue to leverage specialized chips and AI models to accelerate scientific progress, NVIDIA is expanding its presence in scientific computing. Google, for example, uses its Tensor Processing Units (TPUs) and Gemini models to support protein structure research and computational biology. NVIDIA's latest GPU initiative similarly aims to bring AI-powered computing capabilities to scientific and engineering workloads.
The Rubin HPC GPU is part of NVIDIA's Vera Rubin platform and features more than 50,000 CUDA cores, setting a new record for the number of CUDA cores in a single GPU. It also incorporates sparse data acceleration capabilities and NVIDIA's seventh-generation NVLink interconnect technology, which is designed to improve communication speeds between GPUs. These technologies are expected to significantly accelerate simulation workloads, helping researchers and engineers shorten the development cycle from initial exploration to scientific discoveries and industrial design.
Ian Buck, NVIDIA's Vice President of Hyperscale and High-Performance Computing, noted that the growing use of AI agents in scientific research makes faster simulations essential to accelerating discoveries. Rubin HPC is designed to help AI agents evaluate ideas, test hypotheses, and guide subsequent simulation processes more quickly. This could enable scientists and engineers to explore a wider range of possibilities, identify potential development bottlenecks earlier, and accelerate product research and development.