Skip to content
AI & Development News

NVIDIA Rubin HPC Announced, Expected to Ship in 2028

Listen or read this article

5:41

Listen, or read the text only.

NVIDIA Rubin HPC Announced, Expected to Ship in 2028

Kokoro 82M AI-generated voice 6 min read

0:00 5:41

Advertisement

Download audio

File name
nvidia-rubin-hpc-announced-availability-expected-2028-en.mp3
Format
MP3 (audio/mpeg)
Duration
5:41
File size
3.9 MB
Engine
Kokoro 82M

This audio was generated by AI.

You may download and use it freely for personal use.

Advertisement

NVIDIA Rubin HPC Announced, Expected to Ship in 2028

4 min read

NVIDIA Rubin HPC Announced, Expected to Ship in 2028
NVIDIA announced Rubin HPC, a GPU for science and engineering, and said it expects availability in 2028. A single GPU has more than 50,000 CUDA cores, while availability in South Korea and pricing have not been disclosed.
NVIDIA said it expects Rubin HPC to be available in 2028.
According to NVIDIA's announcement, a single Rubin HPC GPU has more than 50,000 CUDA cores.
Rubin HPC uses 7th-generation NVIDIA NVLink for connections between GPUs.
Availability in South Korea, local purchasing channels, pricing, and when orders will open have not been disclosed.
The announcement does not include memory specifications, power consumption, FP64 compute performance, or independent performance test results.
NVIDIA announced Rubin HPC for science and engineering on November 16, 2026. In Korean time, this was 1 a.m. on November 17. It is expected to be available in 2028. NVIDIA did not disclose whether orders can currently be placed.
This article is based on NVIDIA's announcement and was checked as of November 17, 2026.
Rubin HPC announced for scientific and engineering computing
NVIDIA announced Rubin HPC, a GPU for scientific and engineering computing. It uses 7th-generation NVIDIA NVLink for connections between GPUs. This announcement provides an expected availability date and does not mark the start of sales.
Availability expected in 2028, with no schedule disclosed for Korea
NVIDIA said it expects Rubin HPC to be available in 2028. This is the company's estimated schedule. It did not disclose a specific availability date within 2028.
The announcement also does not confirm whether it will be available in Korea. Purchasing channels in Korea have not been disclosed either. It is therefore not yet possible to specify when research institutions or companies in Korea will be able to use it.
Pricing and ordering dates not disclosed
The announcement does not include pricing for Rubin HPC. It also does not disclose when orders can be placed. The expected year of availability alone does not establish that orders or contracts can be placed.
CUDA core and interconnect specifications as announced by NVIDIA
The specifications NVIDIA announced on November 16, 2026, are as follows. The figures in the table are product specifications provided by the company. They are not computing performance figures measured by an independent organization.
Item | NVIDIA's announcement | What the figure represents CUDA core count | More than 50,000 | Number of cores on a single GPU Connections between GPUs | 7th-generation NVIDIA NVLink | Generation of GPU interconnect technology
The CUDA core count is more than 50,000 per GPU. This does not mean exactly 50,000. FP64 computing performance cannot be calculated from this core count alone.
Memory, power consumption, and FP64 performance not confirmed
The announcement does not include detailed memory or power specifications. FP64 computing performance was not provided either. No independent performance measurement results have been confirmed.
Undisclosed item | Further information to confirm Memory | Type, capacity, bandwidth Power | Power consumption Computing performance | FP64 computing performance Independent verification | Performance measurement results obtained independently of the company's announcement
Items to check before considering adoption in Korea
Anyone considering adoption in Korea must first confirm whether the product will be available there. Comparing prices or performance also requires specifications that have not been disclosed. The following items should be checked in NVIDIA's subsequent announcements and product documentation.
· Availability in Korea and purchasing channels in Korea · Specific availability schedule and when orders can be placed · Pricing and memory type, capacity, and bandwidth · Power consumption and FP64 computing performance
Independent performance measurement results should be checked separately from the company's specification announcements. The current announcement alone does not establish the extent of any performance improvement over existing equipment. When the product will be available for purchase in Korea also requires further confirmation.
0:00 0:00
1 / 27

Advertisement

Download text

File name
nvidia-rubin-hpc-announced-availability-expected-2028-en.txt
Format
TXT (text/plain)
Paragraphs
27

Downloads exactly what you see as a text file.

Please cite the source when quoting.

Advertisement

Large text

Makes the text larger and the colors clearer. Turn it on if the text feels too small.

NVIDIA expects Rubin HPC for scientific and engineering computing to be available in 2028.AI-generated image

Key points

  • NVIDIA said it expects Rubin HPC to be available in 2028.
  • According to NVIDIA's announcement, a single Rubin HPC GPU has more than 50,000 CUDA cores.
  • Rubin HPC uses 7th-generation NVIDIA NVLink for connections between GPUs.
  • Availability in South Korea, local purchasing channels, pricing, and when orders will open have not been disclosed.
  • The announcement does not include memory specifications, power consumption, FP64 compute performance, or independent performance test results.

NVIDIA announced Rubin HPC for science and engineering on November 16, 2026. In Korean time, this was 1 a.m. on November 17. It is expected to be available in 2028. NVIDIA did not disclose whether orders can currently be placed.

This article is based on NVIDIA's announcement and was checked as of November 17, 2026.

Rubin HPC announced for scientific and engineering computing

NVIDIA announced Rubin HPC, a GPU for scientific and engineering computing. It uses 7th-generation NVIDIA NVLink for connections between GPUs. This announcement provides an expected availability date and does not mark the start of sales.

Rubin HPC will use seventh-generation NVIDIA NVLink for GPU-to-GPU connectivity.AI-generated image

Availability expected in 2028, with no schedule disclosed for Korea

NVIDIA said it expects Rubin HPC to be available in 2028. This is the company's estimated schedule. It did not disclose a specific availability date within 2028.

The announcement also does not confirm whether it will be available in Korea. Purchasing channels in Korea have not been disclosed either. It is therefore not yet possible to specify when research institutions or companies in Korea will be able to use it.

Pricing and ordering dates not disclosed

The announcement does not include pricing for Rubin HPC. It also does not disclose when orders can be placed. The expected year of availability alone does not establish that orders or contracts can be placed.

CUDA core and interconnect specifications as announced by NVIDIA

The specifications NVIDIA announced on November 16, 2026, are as follows. The figures in the table are product specifications provided by the company. They are not computing performance figures measured by an independent organization.

Item NVIDIA's announcement What the figure represents
CUDA core count More than 50,000 Number of cores on a single GPU
Connections between GPUs 7th-generation NVIDIA NVLink Generation of GPU interconnect technology

The CUDA core count is more than 50,000 per GPU. This does not mean exactly 50,000. FP64 computing performance cannot be calculated from this core count alone.

Memory, power consumption, and FP64 performance not confirmed

The announcement does not include detailed memory or power specifications. FP64 computing performance was not provided either. No independent performance measurement results have been confirmed.

Undisclosed item Further information to confirm
Memory Type, capacity, bandwidth
Power Power consumption
Computing performance FP64 computing performance
Independent verification Performance measurement results obtained independently of the company's announcement

Items to check before considering adoption in Korea

Anyone considering adoption in Korea must first confirm whether the product will be available there. Comparing prices or performance also requires specifications that have not been disclosed. The following items should be checked in NVIDIA's subsequent announcements and product documentation.

  • Availability in Korea and purchasing channels in Korea
  • Specific availability schedule and when orders can be placed
  • Pricing and memory type, capacity, and bandwidth
  • Power consumption and FP64 computing performance

Independent performance measurement results should be checked separately from the company's specification announcements. The current announcement alone does not establish the extent of any performance improvement over existing equipment. When the product will be available for purchase in Korea also requires further confirmation.

Open the link

Opens in a new window.

Advertisement

Sign-in required

Sign in with your Google or Apple account to like, comment, and save highlights.

FAQ

Can Rubin HPC be ordered in Korea now?

The announcement did not disclose whether Rubin HPC would be supplied in Korea or the purchasing channels in Korea. It also did not say when orders would open, so it is not possible to confirm whether it can currently be ordered in Korea.

Will Rubin HPC also be available for use in Korea by 2028?

NVIDIA said it expects to supply it in 2028, but it did not provide a supply schedule for Korea. It also did not disclose exactly when in 2028 it would be supplied.

Can the cost of adopting it in Korea and its scientific computing performance be assessed?

The announcement does not include pricing or power consumption. FP64 computing performance and independent performance test results have not been disclosed either. The announced CUDA core count alone is not enough to assess the cost of adopting it or its actual computing performance.

Sources

Data formats

This content is available in several machine-friendly formats.

Data-only languages (machine translated, files only)

Indonesian JSON MD Portuguese JSON MD Chinese (Traditional) JSON MD Deutsch JSON MD

Verification

This article was drafted with AI and then reviewed and edited by a person.

Reviewed by 신익희 · 편집장 · 2026-11-17

Figures in this article were checked against the source material during generation. · 2026-11-17

This translation has been cross-checked by AI. · 2026-11-17

Reuse & AI usage

Search indexing and AI citation with attribution are welcome. See the license policy for details.

CC BY · License

Loading…

Loading…

Loading…

From Injoys

Request the content you want and take 70% of what it earns

Just leave the subject. We handle production, review, translation and distribution.

See how revenue sharing works

Comments