Nvidia Launches Vera Rubin AI Platform, Announces Full Production With Six TSMC Chips

Nvidia Launches Vera Rubin AI Platform With TSMC Chips

Nvidia Corp. has taken a major leap in AI computing with its new platform, Vera Rubin, featuring six concurrent chips manufactured by Taiwan Semiconductor Manufacturing Co. (TSMC). The company has moved into full production of the platform, marking a strategic departure from its previous approach of releasing only one or two chips at a time.

Full Production Confirmed At CES 2026

CEO Jensen Huang unveiled the Vera Rubin supercomputer platform during his keynote at the 2026 International Consumer Electronics Show (CES) in Las Vegas. “If Vera Rubin is going to be in time for this year, it must be in production by now. And so today I can tell you that Vera Rubin is in full production,” Huang said. This rollout follows a roadmap first hinted at during Huang’s August 2025 visit to Taiwan, where designs were confirmed “taped out” at TSMC.

Multi-Chip Design To Keep Up With Exploding AI Demands

Huang explained the need for multi-chip development by citing the rapid growth of AI. With models growing ten-fold and token generation increasing five-fold annually, developing only one or two chips at a time was no longer feasible. “It is impossible to keep up with those kinds of rates … unless we deploy aggressive extreme co-design, innovating across all the chips, across the entire stack, all at the same time,” he said.

Vera Rubin Platform Components And Architecture

The Vera Rubin suite includes the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet Switch. All components succeed the Blackwell architecture and leverage TSMC’s advanced 3-nanometer process. The flagship system, the Nvidia Vera Rubin NVL72, is a liquid-cooled supercomputer weighing nearly two tons. The platform is named in honour of American astronomer Vera Rubin, whose work provided evidence of dark matter.

Efficiency Gains And Industry Adoption

The new platform offers significant efficiency improvements, including reducing inference costs to one-seventh of Blackwell’s and cutting GPU requirements for Mixture-of-Experts (MoE) model training by 75%. Huang described the platform as “revolutionary,” as all six chips were entirely new designs. Leading AI labs, cloud providers, and system builders, including Amazon Web Services (AWS), Meta, Google, and Microsoft, are expected to be early adopters. Taiwanese firm Hon Hai Precision Industry Co. (Foxconn) is set to manufacture AI servers using the Rubin platform.

(This article has been syndicated from ANI)