AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Nvidia’s Rubin HPC Sets Its Sights On Science With 50,000-Plus Cores on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Nvidia announced Rubin HPC at the SC26 conference on Nov. 16, describing it as a GPU for scientific, engineering and analytics workloads with more than 50,000 CUDA cores on a single chip. The company says the GPU is expected in 2028, but has not disclosed its exact core count, memory specifications, power use, price or independent performance results.

Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on Nov. 16, describing the GPU as designed for scientific, engineering and analytics workloads and claiming it will have more than 50,000 CUDA cores on a single GPU. The chip is expected in 2028, but Nvidia has not yet published the specifications needed to judge its performance or cost for research and high-performance computing systems.

The announcement appeared in a Nvidia blog post by Dan Ernst, the company’s senior director of supercomputing products. Nvidia said Rubin HPC is a new GPU within its Vera Rubin platform, aimed at simulation and data processing. The company did not give an exact core count: its stated figure is “more than 50,000,” which it describes as a record for CUDA cores on one GPU.

Nvidia also cited a sparse data acceleration capability and seventh-generation NVLink connections between GPUs. The company says the design is intended to keep data moving to computing cores, rather than increase compute capacity alone. It did not explain how the sparse-data feature works or provide technical measurements for it in the announcement.

As a rough comparison, Tom’s Hardware reports that the current Rubin GPU has 28,672 CUDA cores, while Nvidia lists 21,760 for its GeForce RTX 5090. Based on the minimum of more than 50,000, that would be at least about 1.7 times the current Rubin count and 2.3 times the RTX 5090 count. Those ratios are arithmetic based on the cited figures, not Nvidia’s own performance comparisons; core counts alone do not establish how quickly a GPU will run a particular workload.

At a glance
announcementWhen: Announced Nov. 16, 2026; expected to be…
The developmentNvidia introduced Rubin HPC, a science-focused GPU it says will have more than 50,000 CUDA cores and is expected in 2028.

Nvidia Rubin HPC · SC26 · Nov 16, 2026

Where a science GPU sits in the AI research loop

An agent suggests ideas, a simulation tests them, the result guides the next step. Nvidia’s pitch: speed up the test.

The loop and one simulation (schematic)

Agentidea Test Refine many candidates next step sooner Inside one simulation: huge dataset GPU cores

Six steps, as Nvidia describes them

  1. Agent ideaAn AI agent proposes candidate materials, molecules or designsNvidia
  2. TestEach idea is checked in a simulation before lab work or a prototypeNvidia
  3. GatherThe code reaches into an immense dataset for a handful of scattered valuesNvidia
  4. ComputeIt calculates a result and repeats this on an enormous scaleNvidia
  5. Feed the coresRubin HPC: sparse data acceleration is meant to keep more than 50,000 CUDA cores supplied with dataNvidia’s claim, not yet measured
  6. RefineFaster results shorten the loop of hypothesis, simulation and refinementNvidia
Expected in 2028. Nvidia has not published FP64 figures, memory, power or a price for Rubin HPC, and there are no independent benchmarks yet.

Source: Nvidia Blog, “NVIDIA Announces Rubin HPC — A New GPU to Accelerate Agents for Science, Engineering and Analytics”, Nov 16, 2026. Schematic, not the real chip or software.

Graphic: Powered by Thorsten Meyer AI

Simulation Capacity for Research

Rubin HPC signals Nvidia’s effort to address workloads where data access and movement can limit performance, including simulations and scientific analysis. Such applications may repeatedly retrieve small pieces of information from large datasets, then calculate and repeat. Nvidia says its design aims to keep the GPU’s cores supplied with data for tasks such as sparse matrix operations, stencil computations, sorting and reductions.

The potential relevance extends beyond conventional supercomputing. Nvidia is also pitching faster simulations as a way to test ideas generated by AI agents working on materials, molecules or designs. If the hardware delivers practical gains on those workloads, researchers and engineers could run more simulation cycles in a given period. That is Nvidia’s rationale, not a demonstrated result: the company has released no independent benchmarks or application-specific performance figures for Rubin HPC.

For buyers planning research systems, the announcement is a signal of a future product direction rather than a basis for procurement decisions. Memory capacity, bandwidth, power demand, system configurations and price can determine whether a GPU fits a workload and budget, and none was disclosed in the source announcement.

Products Worth Considering

DJI Mini 4K Drone Bundle with Strobe Light, Landing Gear, and Propeller Holder Strap

DJI Mini 4K Drone Bundle with Strobe Light, Landing Gear, and Propeller Holder Strap

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

LG 27GS60QC-B Ultragear 27-inch Curved Gaming Monitor (QHD, 180Hz, 1ms, 1000R)

As an affiliate, we earn on qualifying purchases.

How Rubin HPC Fits Nvidia’s Roadmap

Nvidia made the announcement during SC26 in Chicago, scheduled for Nov. 15–20, 2026. Its blog post frames Rubin HPC as a science-focused addition to the Vera Rubin platform. The post’s emphasis is on simulation and data processing, rather than presenting the product primarily as a general AI training accelerator.

The distinction matters because scientific computing can depend on different kinds of arithmetic and data movement than AI model workloads. Nvidia’s technical blog on the Vera Rubin platform reports 33 TFLOPS of native FP64 vector performance for Rubin, compared with 34 TFLOPS for Hopper and 40 TFLOPS for Blackwell. The same table gives higher FP64 matrix figures for newer platforms, but identifies Blackwell and Rubin’s figures as peak performance using Tensor Core-based emulation algorithms. Those platform figures do not establish Rubin HPC’s specifications.

Nvidia’s Jan. 5 platform blog, updated Aug. 21, says dedicated FP64 vector performance remains important for scientific applications that are not dominated by matrix operations. Rubin HPC’s separate announcement does not state its native FP64 performance, so it remains unknown how the new GPU will address that requirement. Tom’s Hardware has suggested the core count could imply a multi-die configuration like Rubin Ultra, but that is the outlet’s inference; Nvidia has not described the chip’s physical layout.

“Designed to accelerate simulation and data processing for science, engineering and analytics workflows.”

— Nvidia, in its Rubin HPC announcement

Specifications Still to Come

The announcement leaves major product and performance questions unanswered. Nvidia has not disclosed Rubin HPC’s exact core count, memory size or bandwidth, power requirements, price, or named systems. It also has not provided FP64 figures for the GPU, an explanation of the sparse-data acceleration feature, or independent benchmarks. Without those details, it is not possible to assess how it will perform across different scientific applications or what infrastructure it will require.

The company’s claim that Rubin HPC has a record number of CUDA cores is attributable to Nvidia, but the post does not explain the comparison set. Nor does the core count by itself show a proportional increase in application performance. The chip’s internal layout is also undisclosed; suggestions that it may use multiple dies are external inference, not a confirmed Nvidia specification.

Details Before the 2028 Release

Nvidia says Rubin HPC is expected to be available in 2028. The announcement does not specify a launch date within that year or identify which systems, research institutions or commercial customers will receive it first.

Further technical disclosures will be needed to evaluate the product: in particular, memory and power specifications, FP64 performance, system configurations and benchmarks on representative scientific workloads. Until Nvidia publishes those details, the 50,000-plus figure describes the announced core count, not a confirmed measure of simulation speed or value.

Key Questions

What is Nvidia Rubin HPC?

Rubin HPC is a GPU Nvidia announced for simulation and data processing in science, engineering and analytics. The company says it is part of the Vera Rubin platform.

How many CUDA cores will Rubin HPC have?

Nvidia says it will have more than 50,000 CUDA cores on one GPU, but has not provided an exact count.

When is Rubin HPC expected to be available?

Nvidia says Rubin HPC is expected in 2028. It has not announced a more specific release date.

Does the core count show how much faster it will be?

No. Core count alone does not establish performance on a specific application. Nvidia has not released independent benchmarks or enough technical specifications to assess the GPU’s speed, power use or cost.

What specifications has Nvidia not disclosed?

The announcement does not give the GPU’s exact core count, memory capacity or bandwidth, power requirements, price or native FP64 performance. It also does not identify systems that will use the chip.

Source: ThorstenMeyerAI.com

You May Also Like

Rendering KIRIN EXPRESS’s Floating-Seam Journey: An Image-Free Method

A fictional 603 km/h maglev journey uses browser-drawn visuals, scroll-linked motion and interactive controls in an AI-built exhibition page.

Jetson Kits and Edge AI Devices: What Beginners Should Expect

Here’s a beginner-friendly overview of Jetson Kits and Edge AI Devices, highlighting essential tips to start your AI journey—find out what you need to know next.

What Developers Should Check When Addressing Broccoli2Shell

CVE-2026-111497 affects Brotli and projects that bundle it. Check dependencies, runtime versions and running processes, then apply available fixes.

Postmortems Are Supposed to Improve Systems—Not Blame People

Keen to turn failures into growth opportunities, learn how postmortems focus on system improvement rather than blame—discover why it matters.