Nvidia Debuts Rubin CPX GPU for 1M-Plus Token Inference
Nvidia announced the Rubin CPX, a new GPU built for inference with context windows larger than 1 million tokens. Part of the Rubin series and aimed at “disaggregated inference” setups, the CPX targets long-context tasks like video generation and coding. Nvidia reported $41.1B in recent data-center sales; the CPX ships end of 2026.