NVIDIA officially unveiled the Vera Rubin supercomputing platform at ISC 2026 recently, targeting advanced computational scenarios such as climate modeling, computational fluid dynamics, and energy exploration. The platform integrates Rubin GPUs with Vera CPUs, leveraging NVLink-C2C, ConnectX-9 SuperNICs, and BlueField-4 DPUs to build an end-to-end accelerated architecture, while employing direct liquid cooling designs to optimize thermal management and energy efficiency. In terms of performance, the Vera Rubin system delivers over 7 exaFLOPS of scientific AI compute power and supports 5 petaFLOPS of native double precision (FP64). Each rack can integrate up to 144 GPUs, enabling overall system performance comparable to TOP500 supercomputers. Jensen Huang, CEO of NVIDIA, stated, "Scientific discovery is racing against the world's most complex challenges. Vera Rubin is a 'rack-scale supercomputer' that unifies simulation, AI, and data processing, accelerating both scientific research and industrial innovation." Currently, several top-tier institutions have announced their adoption of this architecture. The Leibniz Supercomputing Centre’s “Blue Lion” system in Germany will see its computing power increase by approximately 30 times and is expected to go online in 2027. The National Energy Research Scientific Computing Center’s “Doudna” Dell system focuses on molecular dynamics, nuclear fusion, and AI training. Meanwhile, Los Alamos National Laboratory has deployed the “Mission/Vision/Veritas” series, primarily supporting national security, open science, and Agentic R&D. Global manufacturers including HPE, Dell, GIGABYTE, Supermicro, and Bull will simultaneously launch liquid-cooled server racks based on the NVL4 architecture, which are scheduled for official market release in the fourth quarter of this year.