Nvidia Rubin HPC GPU: 50,000 Cores, 2028 Launch & What It Does
Nvidia unveiled Rubin HPC at SC26 in Chicago, a GPU with more than 50,000 CUDA cores built for science and engineering, with systems expected in 2028.
Image: Wikimedia Commons / Geekerwan
CHICAGO — Nvidia unveiled a new graphics processor for scientific computing on Monday, November 16, 2026, at the SC26 supercomputing conference in Chicago, packing more than 50,000 CUDA cores onto a single chip. The company said the Rubin HPC GPU is built for simulation and data processing in science, engineering and analytics, and it expects systems using the chip to be available in 2028.
Rubin HPC is part of Nvidia's Vera Rubin platform, which the company describes as five purpose-built rack-scale systems combined into a multi-rack supercomputer for AI agents doing multi-step reasoning. The chip adds sparse data acceleration, a feature meant to keep its cores supplied with data, and seventh-generation NVLink links for faster communication between GPUs. Nvidia has not published memory capacity, power draw, a price or performance figures for the chip.
Key Takeaways
- Nvidia announced Rubin HPC at SC26 in Chicago on November 16, with more than 50,000 CUDA cores on one GPU.
- It targets simulation and data processing for climate models, engineering, genomics and financial risk.
- Systems are expected in 2028; a 192-CPU Vera rack is due in early 2027.
- Nvidia has not disclosed memory, power, pricing or benchmark results for Rubin HPC.
Nvidia Rubin HPC GPU Specs: 50,000 CUDA Cores, NVLink & 2028 Release
The headline figure is the core count. Nvidia said Rubin HPC carries more than 50,000 CUDA cores on one GPU, a record for the company, and pairs them with sparse data acceleration so the cores are not left waiting on data. Seventh-generation NVLink connects the GPUs to each other. The company also outlined server trays built around the chip: each tray holds two Superchips, each pairing a Vera CPU with two Rubin HPC chips, putting more than 200,000 cores in a single tray. Nvidia gave no memory capacity, power figure, price or performance comparison with the current Rubin GPU.
| Feature | Rubin HPC | What Nvidia said |
|---|---|---|
| CUDA cores | More than 50,000 | Record count, per Nvidia |
| Interconnect | Seventh-generation NVLink | GPU-to-GPU links |
| Special feature | Sparse data acceleration | Keeps cores fed with data |
| Target workloads | Simulation and data processing | Climate, engineering, genomics, finance |
| Availability | 2028 (expected) | No price announced |
What Is the Nvidia Rubin HPC GPU Used For? Science, Climate & Genomics
Nvidia aims Rubin HPC at the patterns behind heavy scientific work: sparse matrix operations, stencil computations, sorting and reductions. Those are the calculations inside climate models, engineering simulations, genomic analysis and financial risk models. The announcement also included Nemotron for Science, an open package for AI-driven research built around a new multimodal model, Nemotron 3.5 Super VL, and an open chemistry dataset of nearly 50,000 reasoning questions. Nvidia said the tools are already in use at the US Department of Energy's Argonne National Laboratory and at RIKEN in Japan. Google's Gemini 4 Hydrogen is also rolling out in Search.
When Will Rubin HPC Be Available? 2028 Date, Vera Racks & Earnings
The chip itself is a long-lead announcement: Nvidia expects Rubin HPC to be available in 2028, and it gave no date for first shipments or a price. The nearer milestone is the Vera CPU rack for high-performance computing, a liquid-cooled rack holding 192 Vera CPUs that Nvidia said is due in early 2027. OpenAI and Anthropic have also lined up 26 million of Google's tensor processing units. The launch is Nvidia's first major product announcement since founder Jensen Huang resigned as chief executive on November 12, with chief financial officer Colette Kress serving as interim chief executive. Nvidia reports third-quarter results on Tuesday, November 17, when investors will seek the first read on demand.
Frequently Asked Questions
When will the Nvidia Rubin HPC GPU be released?
Nvidia expects Rubin HPC in 2028, with no date for first shipments or a price. The nearer milestone is a 192-CPU Vera rack due in early 2027.
How many CUDA cores does the Nvidia Rubin HPC GPU have?
More than 50,000 CUDA cores on a single GPU, which Nvidia calls a record. The chip also uses seventh-generation NVLink and sparse data acceleration to keep those cores supplied with data.
What is the Nvidia Rubin HPC GPU used for?
Simulation and data processing for science, engineering and analytics, including climate models, engineering simulations, genomic analysis and financial risk work.
How much will the Nvidia Rubin HPC GPU cost?
Nvidia has not announced a price, memory capacity, power draw or benchmark results for Rubin HPC.