Las vegas: Jensen Huang, CEO of Nvidia Corp., unveiled the company's latest supercomputer platform at the 2026 International Consumer Electronics Show (CES) in Las Vegas. The platform features six new chips manufactured by Taiwan Semiconductor Manufacturing Co. (TSMC) and is now "in full production."

According to Focus Taiwan, the unveiling of the six chips marks a strategic shift from Nvidia's previous approach of releasing one or two chips at a time. Huang emphasized the necessity of developing multiple chips to keep pace with the rapid growth in AI models and the exponential increase in generated tokens.

Huang stated that the complexity of AI advancements necessitates a comprehensive co-design across all chips simultaneously. This six-chip strategy was initially hinted at in August 2025 during Huang's visit to Taiwan, where he confirmed that new designs had been "taped out" at TSMC.

The Vera Rubin platform, named after astronomer Vera Rubin, introduces a suite of components including the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet Switch. These components replace the Blackwell architecture and are produced using TSMC's advanced 3-nanometer process.

The flagship system of this platform, the Nvidia Vera Rubin NVL72, is a liquid-cooled supercomputer weighing nearly 2 tons. It offers significant cost reductions and efficiency improvements, slashing inference costs to one-seventh of the Blackwell platform and reducing the GPU count needed for training Mixture-of-Experts (MoE) models by 75 percent.

Nvidia announced in a press release that leading AI labs, cloud service providers, and system builders, including Amazon Web Services (AWS), Meta, Google, and Microsoft, are anticipated to be among the first adopters of the new platform. Additionally, Taiwanese company Hon Hai Precision Industry Co. (Foxconn) is reported to be the primary manufacturer of AI servers utilizing Nvidia's Rubin platform.