Nvidia is starting 2026 with an earlier launch of its console new Vera Rubin computing platformfollowing a record year for the Rubin graphics card’s predecessor, Blackwell, fueled by the AI boom (or bubble).
During a press conference leading up to today’s keynote, Dion Harris, senior director of HPC and AI infrastructure solutions at Nvidia, described Vera Rubin as “six chips that make one AI supercomputer.”
These six chips include the Vera CPU, Rubin GPU, NVLink Gen 6 switch, Connect-X9 NIC, BlueField4 DPU, and Spectrum-X 102.4T CPO. The platform will support the third generation confidential data processing and according to Nvidia will be the first trusted rack-scale computing platform.
Nvidia claims that the Rubin GPU is capable of providing five times more AI training computations than Blackwell. The Vera Rubin architecture as a whole can train a vast “mix of experts” (MOE) AI model in the same time as Blackwell, using one-quarter of the GPUs and at one-seventh the token cost.
Rubin was originally scheduled to launch at the end of this year. Its early launch comes today, just months after Nvidia reported record high data center revenues, up 66 percent from a year earlier. This growth was driven by demand for Blackwell and Blackwell Ultra GPUs, which set the bar high for Rubin’s success and served as a driving force for the “AI bubble.” Products and services running on the Rubin platform will be available from Nvidia partners from the second half of 2026.