
NVIDIA is configured to launch a new chip called a Rubin CPX GPU. It is designed specifically for a wide range of AI applications.
The Silicon Valley chip maker announced at the AI Infrastructure Summit in Santa Clara, California.
The new graphical processing unit (GPU) expected to be available at the end of 2026 allows AI systems to manage 1 million tokens of software coding and generated videos with improved speed and efficiency.
Rubin CPX works with the Nvidia Vera CPU and Rubin GPU within the new Nvidia Vera Rubin NVL144 CPX platform.
“We're committed to providing a range of services to our customers,” said Jensen Huang, founder and CEO of Nvidia. “The Vera Rubin platform is taking another leap into the frontier of AI computing, introducing both the next-generation Rubin GPU and a new category of processors called CPX.
“Just RTX revolutionized graphics and physical AI, Rubin CPX is the first Computational Integrated Device Architecture (CUDA) GPU dedicated to large-scale contextual AI modeling the modelling of millions of tokens at once.”
The company said the AI model could require up to 1 million tokens to process an hour of video content. This is difficult for traditional GPUs.
To address this, Rubin CPX is built into a single chip, along with video decoders and encoders and long context inference processing. This integration provides enhanced capabilities for years of applications, such as video search and high-quality generated video.
Built on the Nvidia Rubin Architecture, the Rubin CPX GPU features a cost-effective monolithic die design with powerful NVFP4 computing resources. The GPU offers up to 30 PetaFlops computing with NVFP4 accuracy, providing high performance and accuracy.
Equipped with 128GB of GDDR7 memory, it supports demanding context-based workloads and offers three times faster attention than the NVIDIA GB300 NVL72 system.
Rubin CPX is available in a variety of configurations, including the Vera Rubin NVL144 CPX. According to NVIDIA, the Vera Rubin NVL144 CPX is now able to achieve significant monetization with a potential token revenue of $5 billion for every $100 million invested.
AI companies such as Cursor, Runway and Magic are already investigating the potential of Rubin CPX to power their applications.
Nvidia Rubin CPX is supported by a complete NVIDIA AI stack, from accelerated infrastructure to enterprise-ready software.
In July 2025, Nvidia reached a market capitalization of 4tn amid the AI boom.
