Nvidia has launched Vera Rubin, its next generation of supercomputing hardware, and major partners are already lining up to deploy it.
According to Nvidia's newsroom, a single Vera Rubin rack delivers 7 exaflops of AI performance for science and 5 petaflops of "native FP64" performance — the high-precision math that traditional scientific computing relies on. Nvidia frames this as fitting a TOP500-class supercomputer, normally room-sized, into one rack.
The rollout spans the industry. Server makers Dell and Super Micro both unveiled new systems built around Vera Rubin GPUs, according to MSN's report. Dell said its AI Factory deployments have now passed 5,000, per Stock Titan, and the company added a 144-GPU rack server. Supermicro announced an end-to-end "DCBBS" blueprint built on the Vera Rubin NVL4 design, emphasizing native FP64 for workloads that blend HPC and AI. Investing.com reported that Supermicro's stock rose 8% on the news.
The technology is also headed to one of the U.S. government's premier research sites. Per Nvidia's blog, Los Alamos National Laboratory will build three new supercomputers — named Mission, Vision and Veritas — with HPE and Nvidia, using Nvidia's Vera CPUs and the HPE Cray Supercomputing GX5000 architecture. Nvidia says the goal is "agentic" AI that can accelerate scientific discovery.
Why it matters: Vera Rubin marks the moment AI-optimized chips and traditional scientific supercomputing converge in the same hardware — and the speed at which Dell, Supermicro and a national lab are adopting it signals how quickly that shift is reshaping both enterprise data centers and frontier research.