NVIDIA launches Vera Rubin AI supercomputing platform, partnering with Dell and other vendors

NVIDIA announced its Vera Rubin accelerated computing platform at ISC High Performance 2026, a rack‑scale AI‑enabled supercomputing solution that combines traditional double‑precision performance with large‑scale AI capabilities. The architecture, based on NVIDIA Vera Rubin NVL4, supports up to 144 GPUs per rack, delivers more than 7 exaflops of AI performance and 5 petaflops of FP64 computing, and is targeted at scientific research, climate modelling, quantum chemistry, energy exploration and other high‑performance workloads.

Dell Technologies introduced the PowerEdge XE8812 server, built around the Vera Rubin NVL4 architecture. The server offers 176 CPU cores, up to 144 GPUs per rack, direct liquid cooling and 300 kW power support, and is scheduled for global availability in 2027. Dell highlighted integrated management tools such as iDRAC and PowerRack that aim to reduce deployment time to about six hours.

Penguin Solutions reported earning the NVIDIA AI Factory Specialized Partner designation, underscoring its role in delivering large‑scale AI infrastructure for enterprise and sovereign clients.

VIAVI outlined how Graph Neural Networks and digital twins, accelerated by NVIDIA GPUs, can enable autonomous operation of IP transport networks, emphasizing the need for high‑performance compute to close the validation gap in network changes. NVIDIA also detailed initiatives to help telecom operators adopt AI agents for autonomous network management, citing collaborations with SoftBank, NTT Data and other partners.

The Vera Rubin platform is already slated for deployment in major research centres such as Germany’s Leibniz Supercomputing Centre, the US National Energy Research Scientific Computing Center, and Los Alamos National Laboratory, with broader ecosystem support from Dell, HPE, Bull, GIGABYTE and Supermicro.