Nvidia has launched a new Rubin chip architecture

Today at Consumer Electronics ShowNvidia CEO Jensen Huang officially launched the company’s new Rubin computing architecture, which he described as the latest in AI hardware. The new infrastructure is currently under construction and is expected to be completed in the second half of the year.

“Vera Rubin was created to solve this big problem we have: The amount of computation required for AI is increasing.” Huang told the audience. “Today, I can tell you that Vera Rubin is making everything.”

Rubin Architecture, first announced in 2024it is the latest result of Nvidia’s hardware development movement, which has turned Nvidia into the most valuable company in the world. Rubin’s architecture will replace Blackwell’s architecture, which replaced Hopper’s and Lovelace’s architecture.

Rubin’s chips are already designed for use by nearly every major cloud provider, including top partners Nvidia and Anthropic, OpenAIand Amazon Web Services. Rubin systems will also be used HPE’s Blue Lion supercomputer and Doudna’s most advanced machine at Lawrence Berkeley National Lab.

It is called astronomer Vera Florence Cooper RubinRubin’s architecture consists of small chips designed for concert use. The Rubin GPU stands at the center, but the architecture also addresses the growing challenges of storage and connectivity with new innovations in Bluefield and NVLink, systems respectively. The architecture also includes the new Vera CPU, designed for aging.

Explaining the benefits of the new storage, Nvidia’s director of AI solutions Dion Harris also pointed to the growing cache memory for today’s AI systems.

“When you start supporting new types of workflows, like AI-assisted or long-term tasks, that puts a lot of stress and demand on your KV cache,” Harris told reporters on a call. memory system used by AI models to reduce input. “That’s why we introduced a new storage unit that connects externally to the computing device, which allows you to expand your storage pool even more efficiently.”

As expected, the new architecture also represents a major advance in speed and power efficiency. According to Nvidia’s tests, Rubin’s architecture will work three and a half times faster than Blackwell’s previous architecture for model training tasks and five times faster for display tasks, up to 50 petaflops. The new platform will also support eight times more power per watt.

Rubin’s new capabilities come amid intense competition to build AI infrastructure, which has seen AI labs and cloud providers push Nvidia’s chips and the hardware needed to power them. Calling the money in October 2025, Huang estimated between $3 trillion and $4 trillion will be used for AI infrastructure over the next five years.

Follow along with all of TechCrunch Coverage of the annual CES conference here.