GPUDatacenterReleased —

NVIDIA Rubin CPX

128 GB GDDR7

Rubin CPX is a data-center GPU that NVIDIA announced on September 9, 2025 for the context (prefill) phase of massive-context AI inference, such as million-token code assistance and long-form video generation. It uses a monolithic die with 128 GB of GDDR7 memory and on-chip video encoders and decoders, and was designed to work alongside Rubin GPUs and Vera CPUs in the Vera Rubin NVL144 CPX rack. NVIDIA originally targeted availability for the end of 2026, but the product did not appear on its GTC 2026 roadmap in March 2026.

General

Manufacturer: NVIDIA
Model: Rubin CPX
Segment: Datacenter
GPU chip: Rubin CPX (monolithic)
Architecture: Rubin
Release date: —
Price: Search on Amazon

Memory

Memory size: 128 GB
Memory type: GDDR7
Bus width: —
Bandwidth: —
Memory clock: —

Clocks & silicon

GPU clock: —
Boost clock: —
Shader units: —
Tensor / AI cores: —
RT cores: —
Transistors: —
Technology (nm): —
TDP / TBP: —

Interface & APIs

Bus interface: —
SLI / multi-GPU: —
DirectX: —
OpenGL: —
Vulkan: —
Other APIs: —

AI support

Precisions: FP8FP16BF16INT8INT4TF32FP64
AI accelerator: Tensor Cores with NVFP4 and attention acceleration
Peak AI TOPS: 30,000 TOPS
Notes: Up to 30 petaflops of NVFP4 compute, with attention acceleration that NVIDIA rated at three times that of GB300 NVL72 systems.

All NVIDIA GPUs →

Datacenter gpus from NVIDIA announced around the same time as the NVIDIA Rubin CPX. Add any of them to your compare list to see the specs side by side.

gpu