Nvidia announces Rubin HPC GPU with 'record number of CUDA cores' at Supercomputing 2026 — new chip accelerates agentic HPC workloads, arrives in 2028 - Tom's Hardware
Nvidia has announced its new Rubin HPC GPU featuring a record number of CUDA cores at Supercomputing 2026, designed to accelerate agentic HPC workloads.
The Rubin HPC GPU is positioned to significantly enhance performance for high-performance computing applications, particularly those involving agentic workloads that require sophisticated decision-making capabilities.
While specific technical details about the number of CUDA cores and other specifications were not disclosed in the announcement, the emphasis on a 'record number' suggests substantial performance improvements over previous generations.
Future Impact
The 2028 release timeline indicates this is part of Nvidia's long-term roadmap for advancing computing technology, with potential applications in scientific research, artificial intelligence development, and complex simulations.