Nvidia on 16 November 2026 announced Rubin HPC, a new graphics processing unit (GPU) built for simulation and data processing in science, engineering and analytics. The company made the announcement at the SC26 supercomputing conference and said the chip is expected to be available in 2028.

In a company blog post, Nvidia said Rubin HPC is part of its Vera Rubin platform and carries more than 50,000 CUDA cores on a single GPU, which it described as a record number.

What the chip is designed to do

Nvidia said the GPU pairs that core count with two other features: a sparse data acceleration capability meant to keep the cores supplied with data, and seventh-generation NVLink, the company's interconnect technology for linking one GPU to another.

The company explained the reasoning this way. Workloads such as climate models, engineering simulations, genomic analysis and financial risk calculations tend to follow one pattern: a program reaches into a very large dataset, pulls out a few scattered values, computes a result and repeats the process on an enormous scale. Nvidia said Rubin HPC is therefore designed to scale data movement and computing power together, so that the cores "spend their time computing rather than waiting".

The link to AI agents

Nvidia is pitching the chip at a way of working in which AI agents help researchers and engineers sift through candidate materials, molecules and designs. Those ideas still have to be tested through simulation, and the company argues that demand for GPU-accelerated simulation will rise as a result.

"For agents tackling science and engineering, faster simulation means faster discovery," said Ian Buck, vice president of hyperscale and high-performance computing at Nvidia, in the company's post.

Nvidia also said its CUDA-X software libraries offer optimised versions of complex algorithms that agents can use together with the new chip.

What is not yet known

The announcement is a preview rather than a product launch. Nvidia's post gives no price, no performance benchmarks against its existing chips and no list of customers or supercomputing centres that plan to use Rubin HPC. The speed-up claims are the company's own and have not been independently tested.

What happens next

Nvidia said Rubin HPC is expected to be available in 2028. The company published several other announcements from SC26 on the same day, including a desktop developer kit for scientific computing.

PS · Oleg Alexandrov, CC BY-SA 4.0