NVIDIA releases Vera Rubin platform: a single-rack supercomputer will deliver 7 Exaflops of AI computing power

BlockBeats News, June 22 — NVIDIA announced the launch of the Vera Rubin supercomputing platform, targeting scientific computing and high-performance computing (HPC). The single-rack system can deliver over 7 Exaflops of AI computing power and 5 Petaflops of FP64 double-precision performance, officially described as "a single rack capable of achieving TOP500-level supercomputing capability."

This platform integrates Rubin GPU and Vera CPU architectures, combined with technologies such as NVLink-C2C, ConnectX-9 SuperNIC, and BlueField-4 DPU, supporting unified computing needs for AI training, scientific simulation, and data-intensive research.

NVIDIA states that Vera Rubin will focus on high-complexity scientific computing tasks such as climate modeling, computational fluid dynamics, quantum chemistry, and energy exploration, and will support an integrated workflow of "AI + simulation + data analysis."

At the application level, institutions like the Leibniz Supercomputing Centre, the U.S. Department of Energy's National Energy Research Scientific Computing Center (NERSC), and Los Alamos National Laboratory have plans to adopt this platform to build next-generation supercomputing systems.

Additionally, vendors such as Bull, Dell Technologies, HPE, and Supermicro will also launch high-density liquid-cooled rack-mounted supercomputing solutions based on Vera Rubin. NVIDIA CEO Jensen Huang stated that the platform aims to merge AI, simulation, and data processing into "a new scientific tool" to accelerate scientific discovery and industrial innovation.