NVIDIA Unveils New GPU for Long-Context AI Inference

Nvidia Unveils Rubin CPX GPU for Extended Context Windows

During Tuesday’s AI Infrastructure Summit, Nvidia revealed its latest GPU, the Rubin CPX. This new processor is specifically engineered to handle context windows exceeding 1 million tokens.

The CPX is a key component of Nvidia’s upcoming Rubin series. It’s been optimized for the efficient processing of extensive sequential data.

Disaggregated Inference and Performance Gains

Nvidia intends for the Rubin CPX to function within a wider “disaggregated inference” framework. This approach aims to deliver enhanced performance for tasks requiring substantial context, such as video creation and software engineering.

By utilizing a disaggregated infrastructure, users can expect improvements in handling complex, long-form data.

Nvidia’s Financial Success

The company’s consistent innovation in GPU technology has translated into significant financial gains. Nvidia recently reported $41.1 billion in revenue from its data center business for the last quarter.

Data center sales represent a major growth area for Nvidia, fueled by the increasing demand for AI processing power.

Availability

The Rubin CPX is currently scheduled for release in late 2026.

This timeline allows for further refinement and integration into existing and emerging AI infrastructure solutions.