NVIDIA announced on July 21 (local time) that it has officially begun supplying its next-generation artificial intelligence (AI) platform "Vera Rubin" (VR), which follows its "Grace Blackwell" (GB) platform.

AFP Yonhap News

AFP Yonhap News

View original image

Vera Rubin consists of seven new chips, including NVIDIA’s "Rubin" graphics processing unit (GPU) and "Vera" central processing unit (CPU). Its core system, NVL72, is an AI computing system that integrates 72 Rubin GPUs and 36 Vera CPUs into a single rack.


Vera Rubin is designed to optimize every part of the system—from semiconductors to data center power infrastructure—in order to enhance power efficiency and reduce token processing costs. According to the company, CoreWeave recently conducted the first benchmark on the Vera Rubin platform using the "DeepSeek-R1" model. The results showed that tokens processed per second per megawatt (MW) were ten times higher than with the previous Grace Blackwell NVL72 system. NVIDIA highlighted that this achievement dramatically boosts power efficiency—a key competitive factor for AI data centers where power supply is limited. In addition, the company emphasized that by designing the coolant supply temperature for new AI data centers at 45 degrees Celsius and implementing a closed-loop liquid cooling system, they can save millions of gallons of water annually per megawatt of power capacity.


NVIDIA stated that Vera Rubin boasts the largest-ever rack-scale supply chain, spanning over 350 production facilities across 30 countries. Currently, Vera Rubin NVL72 AI racks are being supplied and operated by Google Cloud, Microsoft’s Azure cloud platform, Oracle Cloud, and CoreWeave, among others. During a briefing at the company’s California headquarters, NVIDIA Vice President Ian Buck said, "Vera Rubin is now in full-scale production, and all major clients are building out their systems."


He also took the unusual step of directly comparing products with competitors. The company revealed that the new Vera CPU executes Python programming language code up to 1.8 times faster than AMD’s "Turin" design. However, AMD is soon expected to officially unveil its next-generation server CPU, "Venice," as the successor to Turin.



The main reason behind NVIDIA’s aggressive product promotion is the intensifying competition in the AI chip market. According to Bloomberg News, NVIDIA faces a situation where it must prove both the timely supply and superior performance of its next-generation products over its competitors. Recently, AMD announced that Microsoft will deploy its rack-scale AI system "Helios" on Azure. In addition, Broadcom is working to expand its market share, while major clients such as Amazon are also pursuing the development of their own AI chips, further intensifying the competition.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing