Nvidia launches Vera Rubin NVL72 AI supercomputer at CES — promises up to 5x greater inference performance and 10x lower cost per token than Blackwell, coming 2H 2026

Nvidia unveiled the Vera Rubin NVL72 AI supercomputer at CES, boasting up to 5x greater inference performance and 10x lower cost per token compared to the Blackwell system. The supercomputer features a new AI data center rack-scale architecture with six types of chips, including the Vera CPU and Rubin GPU. With 50 PFLOPS of inference performance and 35 PFLOPS of training performance per Rubin GPU, the system aims to meet the growing demand for AI compute. Nvidia also introduced NVLink 6 for enhanced networking capabilities and highlighted improvements in reliability, availability, and serviceability. The company plans to start volume production of Vera Rubin NVL72 systems in the second half of 2026.