Nvidia wants to own every chip in AI data centers

Nvidia is going crazy this week unveiled its novel Vera Rubin chip, revealing novel performance tests for its GPU and CPU combination ahead of rival AMD’s annual product event in San Francisco on Thursday.

Last week, during a lengthy technical workshop at the company’s headquarters in Santa Clara, California, Nvidia executives showed off the chip’s increased power and performance to a petite group of journalists. The most significant conclusion: Nvidia, which has long specialized in the production of graphics processors, is increasingly trying to position itself as a supplier of processors that can power AI agents.

While GPUs are still the primary hardware companies utilize to train and run AI models, the industry’s shift toward more elaborate, agent-based systems has increased the need for processors that can coordinate data flows, networking, and other software tasks. This is probably one of the reasons why Nvidia is so eager to promote itself as a supplier of complete AI systems, not just AI chips.

Vera Rubin is Nvidia’s successor to its Grace Blackwell hybrid superchip system and is the linchpin of its immediate future powering the artificial intelligence industry. It is designed to offer one CPU for every two GPUs. In one Vera Rubin NVL 72 superchip system, there are 36 Vera GPUs for every 72 Rubin GPUs. Nvidia also sells the Vera processor as a standalone product, and indeed it does he reportedly told Chinese customers they may be ready in August.

Nvidia executives stressed that the novel Vera Rubin NVL72 racks – a stack of chips housed in a single liquid-cooled platform – are much more plug-and-play than some of the company’s previous products. During a brief tour of Nvidia’s Silicon Valley data center lab, Nvidia executives said OpenAI already has one Vera Rubin rack in utilize.

Nvidia CEO Jensen Huang didn’t show up at a workshop in Santa Clara last week; was in Japan announcing the chipmaker’s novel product cooperation with many Japanese companies to develop artificial intelligence for robotics. Instead, the briefing was led by Ian Buck, Nvidia’s longtime vice president of accelerated computing and the architect behind the company’s CUDA software.

“We are well on our way to developing new architectures, not only for GPUs but also for CPUs,” Buck told reporters. “We will continue to innovate because it’s either you do it or you die in Silicon Valley.”

The meetings were held in Huang’s executive briefing center, where many nearby desks were littered with bags of Taiwanese snacks that the CEO had brought back from a recent trip to Computex, the massive annual semiconductor trade show in Taipei, an Nvidia spokesman told WIRED.