Taipei [Taiwan], January 7: Nvidia has moved into full production of its latest AI computing platform, Vera Rubin, featuring six concurrent chips manufactured by Taiwan Semiconductor Manufacturing Co. (TSMC). The announcement was made by CEO Jensen Huang during his keynote at CES 2026 in Las Vegas.
Vera Rubin marks a shift from Nvidia’s previous development cycle, which typically released one or two chips at a time. Huang explained that the multi-chip approach is necessary to keep pace with rapidly growing AI workloads, where models and token generation are expanding exponentially.
The platform includes the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet Switch, all built on TSMC’s 3-nanometer process. Its flagship system, the Vera Rubin NVL72, is a liquid-cooled supercomputer weighing nearly two tons. The platform is named in honor of American astronomer Vera Rubin, known for her work on dark matter.
Efficiency gains with Vera Rubin include reducing inference costs to one-seventh of Nvidia’s Blackwell platform and cutting the GPU count required for training Mixture-of-Experts (MoE) models by 75%. Leading AI labs, cloud providers, and system builders—including Amazon Web Services, Meta, Google, and Microsoft—are expected to adopt the platform first. Hon Hai Precision Industry Co. (Foxconn) will serve as the primary manufacturer of servers built on the Rubin platform.
About The Author
