Nvidia is ready to launch a brand new chip, known as Rubin CPX GPU, designed particularly for in depth AI functions.
The Silicon Valley chipmaker made the announcement on the AI Infra Summit in Santa Clara, California.
This new graphical processing unit (GPU), anticipated to develop into accessible on the finish of 2026, will allow AI methods to handle million-token software program coding and generative video with improved pace and effectivity.
The Rubin CPX operates at the side of NVIDIA Vera CPUs and Rubin GPUs throughout the new NVIDIA Vera Rubin NVL144 CPX platform.
Nvidia founder and CEO Jensen Huang mentioned: “The Vera Rubin platform will mark one other leap within the frontier of AI computing — introducing each the next-generation Rubin GPU and a brand new class of processors known as CPX.
“Simply as RTX revolutionised graphics and bodily AI, Rubin CPX is the primary Compute Unified Machine Structure (CUDA) GPU purpose-built for massive-context AI, the place fashions motive throughout tens of millions of tokens of information directly.”
The corporate mentioned that AI fashions could require as much as 1 million tokens to course of an hour of video content material, which is difficult for conventional GPUs.
To deal with this, the Rubin CPX incorporates video decoders and encoders, together with long-context inference processing, right into a single chip. This integration offers enhanced capabilities for long-format functions, together with video search and high-quality generative video.
Constructed on the NVIDIA Rubin structure, the Rubin CPX GPU includes a cost-efficient monolithic die design with highly effective NVFP4 computing sources. The GPU delivers as much as 30 petaflops of compute with NVFP4 precision, offering excessive efficiency and accuracy.
It’s geared up with 128GB of GDDR7 reminiscence to assist demanding context-based workloads and gives 3 times quicker consideration capabilities in comparison with the NVIDIA GB300 NVL72 methods.
The Rubin CPX can be accessible in numerous configurations, together with the Vera Rubin NVL144 CPX. Based on Nvidia, the Vera Rubin NVL144 CPX permits corporations to realize important monetisation, with potential token income of $5bn for each $100m invested.
AI corporations, reminiscent of Cursor, Runway and Magic, are already exploring the potential of Rubin CPX to boost their functions.
Nvidia Rubin CPX can be supported by the whole NVIDIA AI stack, starting from accelerated infrastructure to enterprise-ready software program.
In July 2025, Nvidia reached a market capitalisation of $4tn amid the AI growth.
Join our each day information round-up!
Give your corporation an edge with our main trade insights.
