Nvidia CEO Jensen Huang introduced the company’s next-generation AI computing platform, Vera Rubin, at CES 2026, saying it is now in full production. The system features six new chips manufactured by Taiwan Semiconductor Manufacturing Co.
The rollout marks a shift from Nvidia’s past one- or two-chip cycles, reflecting the rapid growth of AI models and data generation. Huang said aggressive, system-wide co-design is needed to keep pace with the industry’s expansion.
The platform includes the Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet switch, built mainly on TSMC’s advanced 3-nanometer process. It succeeds the Blackwell architecture.
The flagship Vera Rubin NVL72 is a liquid-cooled supercomputer designed to cut inference costs to one-seventh of Blackwell and reduce GPU needs for certain AI training by 75%. Major cloud providers including AWS, Meta, Google, and Microsoft, are expected to adopt the platform, with Foxconn serving as a key server manufacturer.
