Be first to deploy NVIDIA Vera Rubin

The NVIDIA Vera Rubin platform is the next generation of AI compute architecture from NVIDIA. It brings together Rubin GPUs, Vera CPUs, networking and system scale components into a unified AI infrastructure designed for training, inference and reasoning workloads.

The Rubin GPU architecture provides high performance compute for AI models, with cores and interconnects optimized to support large models and long context reasoning. It is designed to deliver greater throughput for training, inference and generative tasks compared with prior NVIDIA GPU generations.

The Vera CPU is built for AI first workloads and is tightly integrated with Rubin GPUs using high speed links. It supports data intensive processing and enables teams to run complex workloads at rack scale while maximizing the performance of the Rubin GPU compute.

NVL72 refers to a rack scale configuration within the Vera Rubin platform that combines 72 Rubin GPUs with 36 Vera CPUs and high performance networking components. It is one of the first confirmed configurations designed for large scale training and inference.

NVIDIA has indicated that Rubin based systems are in production and are expected to be available through partners and service providers in the second half of 2026, with early configurations such as NVL72 among the first to enter the market.

DGX systems are fully integrated AI platforms from NVIDIA, ready to use out of the box. HGX provides the same GPU and interconnect technology but lets partners and OEMs build custom configurations. With Rubin, both DGX and HGX can deploy Rubin GPUs and Vera CPUs depending on whether you want a turnkey system or a flexible setup.