NVIDIA Announces Rubin HPC GPU With Over 50,000 CUDA Cores
NVIDIA officially announced the Rubin HPC GPU at the SC26 conference on 17 November 2026, targeting artificial intelligence, datacentre, and scientific computing workloads rather than consumer PCs. Part of the Vera Rubin platform, the accelerator packs more than 50,000 CUDA cores on a single GPU alongside seventh-generation NVIDIA NVLink interconnect technology. NVIDIA expects the Rubin HPC to become available from 2028 onwards, with pricing yet to be disclosed. For Malaysian enterprises, universities, and AI datacentre operators, the announcement outlines future enterprise compute hardware, though local availability and Ringgit pricing remain unconfirmed.
Key Facts
The NVIDIA Rubin HPC is a datacentre graphics accelerator announced at SC26 on 17 November 2026 for high-performance computing and AI training. It integrates over 50,000 CUDA cores and seventh-generation NVIDIA NVLink within the Vera Rubin ecosystem, with commercial availability expected from 2028 onwards and no pricing announced yet.
| Attribute | Value |
| Product Name | NVIDIA Rubin HPC |
| Platform | Vera Rubin |
| Announcement Event and Date | SC26, 17 November 2026 |
| CUDA Cores | More than 50,000 on a single GPU |
| Interconnect | Seventh-generation NVIDIA NVLink |
| Key Acceleration Feature | Sparse data acceleration capability |
| Target Workloads | Science, engineering, analytics, AI training, and datacentre simulation |
| Expected Availability | From 2028 onwards |
| Price | Not disclosed by NVIDIA |
NVIDIA expects the Rubin HPC GPU to enter availability from 2028 onwards with more than 50,000 CUDA cores on a single processor.
What Hardware Features Does the Rubin HPC GPU Include?
The Rubin HPC GPU combines more than 50,000 CUDA cores on a single chip with sparse data acceleration and seventh-generation NVIDIA NVLink interconnect technology. NVIDIA states that this balanced architecture scales data movement alongside raw compute so that cores stay supplied with data rather than waiting idly during intensive tasks.
According to NVIDIA, scaling compute alone is insufficient for scientific computing if memory and data transfers bottleneck execution. By pairing sparse data acceleration with high-speed GPU-to-GPU connectivity, the Rubin HPC targets recurring simulation patterns such as sparse matrix operations, stencil computations, sorting, and reductions. Full technical figures covering clock speeds, memory capacity, and thermal design power were not disclosed in the announcement.
The Rubin HPC pairs over 50,000 CUDA cores with sparse data acceleration and seventh-generation NVIDIA NVLink to accelerate matrix, stencil, sorting, and reduction workloads.
How Does Rubin HPC Fit Into the Vera Rubin Platform?
Rubin HPC joins the broader Vera Rubin hyperscaler Agentic AI platform, which NVIDIA originally unveiled at GTC 2026. The complete Vera Rubin architecture integrates the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, an NVIDIA Spectrum-6 Ethernet switch, and a newly integrated Groq 3 LPU.
- A Vera Rubin compute configuration houses 72 Rubin GPUs and 36 Vera CPUs linked via NVLink 6, alongside ConnectX-9 SuperNICs and BlueField-4 DPUs.
- NVIDIA also detailed a dedicated Vera CPU Rack housing 256 units of the Vera CPU, liquid-cooled using an MGX infrastructure.
- The source announcement does not specify whether Rubin HPC configurations alter the 72-GPU and 36-CPU ratio used in the standard Vera Rubin setup.
NVIDIA's Vera Rubin system links 72 Rubin GPUs and 36 Vera CPUs via NVLink 6, complemented by a liquid-cooled 256-unit Vera CPU Rack built on MGX infrastructure.
Why Is NVIDIA Targeting Scientific Simulation and Agentic AI?
NVIDIA designed the Rubin HPC specifically for AI training, datacentres, academia, and scientific research rather than mass-market consumer graphics. The company positions the processor to speed up science, engineering, and analytics simulations that AI agents use to test ideas and determine their next steps.
For agents tackling science and engineering, faster simulation means faster discovery. Rubin HPC accelerates the simulations agents use to test ideas and guide their next steps, helping scientists and engineers explore more possibilities and bring breakthroughs closer.
Ian Buck, Vice President of Hyperscale and High-Performance Computing at NVIDIA
NVIDIA positions the Rubin HPC exclusively for datacentre AI training, academic research, and scientific simulation rather than consumer computing.
What It Means for Malaysian Buyers
No Malaysian pricing, local launch date, authorised distributor, or datacentre deployment partner has been confirmed for the NVIDIA Rubin HPC. Because the accelerator targets hyperscale datacentres and research institutions and will not ship until 2028 onwards, general Malaysian PC buyers and gamers will see no direct consumer hardware from this release.
For Malaysian universities, national research labs, and enterprise cloud providers planning multi-year infrastructure upgrades, the 2028 timeline serves only as an early roadmap marker. Organisations evaluating near-term AI or high-performance computing clusters should continue working with currently available hardware, as NVIDIA has provided neither unit pricing nor full power and cooling specifications for local procurement planning.
NVIDIA has confirmed neither Ringgit pricing nor Malaysian distribution details for the Rubin HPC ahead of its expected 2028 availability window.
Common Questions
Readers evaluating NVIDIA's SC26 announcement often ask when the Rubin HPC GPU will reach market, whether it serves consumer PCs, and how it relates to the Vera Rubin system. The three answers below summarise every confirmed detail from NVIDIA's announcement alongside what remains undisclosed.
When will the NVIDIA Rubin HPC GPU be available?
NVIDIA expects the Rubin HPC GPU to become available from 2028 onwards, following its initial unveiling at SC26 on 17 November 2026. The company has not confirmed a specific quarter, regional rollout schedule, or Malaysian availability date, nor has it published official unit pricing.
Is the Rubin HPC GPU meant for gaming or consumer PCs?
No, NVIDIA states the Rubin HPC is not a GPU for mass-market consumers. It is engineered specifically for datacentres, artificial intelligence training, scientific simulation, engineering, analytics, and academic research, featuring over 50,000 CUDA cores and seventh-generation NVLink interconnect technology for multi-GPU server clusters.
What hardware is included in the Vera Rubin platform?
First announced at GTC 2026, the Vera Rubin hyperscaler Agentic AI platform combines 72 Rubin GPUs and 36 Vera CPUs connected via NVLink 6, alongside ConnectX-9 SuperNICs, BlueField-4 DPUs, an NVIDIA Spectrum-6 Ethernet switch, a Groq 3 LPU, and a dedicated 256-unit Vera CPU Rack.
All confirmed specifications for the Rubin HPC originate from NVIDIA's SC26 announcement, while pricing and local availability remain undisclosed.
Sources and Methodology
This dispatch is based on the Lowyat.NET report published and checked on 17 November 2026, covering NVIDIA's official SC26 announcement. No currency conversions or Ringgit estimates were applied because NVIDIA did not disclose pricing for the Rubin HPC GPU or the Vera Rubin platform.
- Primary Source: Lowyat.NET (NVIDIA Announces Rubin HPC GPU At SC26, by John Law, citing NVIDIA), checked on 17 November 2026.
- Currency Methodology: No USD-to-RM conversions were performed as official hardware pricing has not been released.
All specifications, quotes, and release timelines in this report were verified against the 17 November 2026 Lowyat.NET report citing NVIDIA.
Related coverage
More Laptops & PCs →Filed under
- #Tech
- #Nintendo
- #PC Gaming
Network · latest from sister sites