At the SC26 conference, Nvidia introduced the Rubin HPC processor, featuring more than 50,000 CUDA cores, and announced extensions to its Vera CPU family to boost high‑performance computing workloads.
Nvidia unveiled its new Rubin HPC GPU and expanded the Vera CPU family at the Supercomputing 2026 (SC26) conference, signaling a major push into high‑performance computing (HPC) alongside its AI dominance.
Rubin HPC GPU: Scale and Architecture
The Rubin processor boasts more than 50,000 CUDA cores, a memory bandwidth exceeding 2 TB/s, and a custom interconnect designed for low‑latency scaling across thousands of nodes. Nvidia says the chip targets scientific simulations, weather modeling, and large‑scale AI training workloads that demand both raw compute and efficient data movement.
Vera CPU Lineup Extension
Alongside Rubin, Nvidia announced new variants of its Vera CPU family, adding higher core counts and support for the latest DDR5 and PCIe 5.0 standards. The extensions aim to provide a unified software stack for developers who want to run CPU‑intensive tasks and GPU‑accelerated kernels on the same platform.
- Up to 128 cores per Vera socket
- Integrated support for Nvidia’s CUDA and cuBLAS libraries
- Optimized power envelope for dense rack deployments
Implications for the HPC Market
Analysts view Nvidia’s dual‑track strategy as an effort to capture more of the traditional HPC market, which has long been dominated by x86 and ARM vendors. By offering both a massive GPU and a complementary CPU line, Nvidia hopes to simplify system design and reduce the software integration burden for research institutions.
“The combination of Rubin and the new Vera CPUs gives us a single‑vendor solution that can handle the most demanding simulations without the usual hardware juggling,” said a senior engineer at a leading national lab.
For readers interested in broader infrastructure trends, see the Infrastructure category on uxc.news.
Related coverage: about Nvidia’s roadmap on its Nvidia Corporation page on Momentum.
Network World coverage of Nvidia expands AI infrastructure push into high‑performance computing