Nvidia unveils Vera Rubin supercomputing platform at ISC 2026

Nvidia unveiled its Vera Rubin platform Monday at ISC 2026 in Hamburg, saying one rack delivers over 7 exaflops of AI performance.Investing.com
Major deployments include Germany's Blue Lion, Berkeley Lab's Doudna, and two Los Alamos systems for nuclear stewardship and open science.Tom's Hardware+1
Supermicro, Dell , and HPE plan to ship liquid-cooled Vera Rubin systems by Q4 2026.Investing.com+1

NVIDIA unveiled its Vera Rubin accelerated computing platform for scientific supercomputing on Monday at ISC High Performance 2026 in Hamburg, Germany, positioning a single rack of the new system as capable of rivaling machines on the TOP500 list of the world's most powerful supercomputers.Investing.com+1

The company said a Vera Rubin system can deliver more than 7 exaflops of AI performance and 5 petaflops of native double-precision (FP64) computing with up to 144 GPUs per rack, combining traditional scientific simulation with AI capabilities for workloads including climate modeling, computational fluid dynamics, and energy exploration.Investing.com+1

Major Science Deployments

Several leading research institutions have committed to Vera Rubin-based systems. Germany's Leibniz Supercomputing Centre is building Blue Lion, a supercomputer offering roughly 30 times more computing power than the current SuperMUC-NG system, built by HPE using next-generation Cray technology. The National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory will deploy Doudna, a Dell Technologies system named for Nobel laureate Jennifer Doudna, expected to serve over 11,000 researchers when it launches and deliver at least 10 times the performance of its predecessor Perlmutter. Los Alamos National Laboratory has contracted with HPE Hewlett Packard Enterprise Company to build its Mission and Vision supercomputers, with Mission serving nuclear stockpile stewardship and Vision continuing open science and AI research, both scheduled for 2027.Tom's Hardware+2

Industry Partners Scale the Platform

Super Micro Computer announced a Data Center Building Block Solutions Blueprint for the NVIDIA Vera Rubin NVL4 at ISC 2026, designed to help research institutions deploy converged HPC and AI infrastructure at scale. Each 3.2-megawatt Scalable Unit contains eight liquid-cooled compute racks, housing up to 1,152 NVIDIA Rubin GPUs and 576 Vera CPUs connected via Quantum-X800 InfiniBand networking.Investing.com+1

Manufacturers including Dell, Gigabyte, HPE, and Supermicro have announced plans to offer liquid-cooled Vera Rubin systems, with NVL4-based configurations expected to be available in the fourth quarter of 2026.Investing.com

A Platform Built for Converged Science

The Vera Rubin platform integrates six co-designed chips — the Rubin GPU, Vera CPU, NVLink 6, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X switch — in a liquid-cooled architecture. First revealed at CES 2026 in January, the platform was designed from the outset to unify simulation, data analytics, and AI within a single system, marking what NVIDIA describes as a new phase in scientific computing where AI and traditional simulation reinforce each other.The Verge+3

Sources (30)