[REVISION HISTORY]
Supermicro and NVIDIA expand Vera Rubin AI infrastructure
6 recorded revisions of this situation.
What changed
2026-11-17 13:28 UTC → 2026-11-18 15:22 UTC ·
added
removed
Super Micro Computer, Inc. has commenced shipments of NVIDIA Vera Rubin NVL72 racks, integrating its Data Center Building Block Solutions with DLC-2 direct liquid cooling technology. These systems, designed to manage high thermal loads, feature 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs per rack. To standardize industry infrastructure, NVIDIA has opened its Vera Rubin AI platform rack and compute tray designs to the industry via the third-generation MGX open architecture. This initiative utilizes the MGX 1.2 specification to support rack deployments up to 250kW, featuring blind-mate liquid cooling and liquid-cooled power busbars. Supermicro has expanded its portfolio with a liquid-cooled CPU rack containing 256 NVIDIA Vera CPUs tailored for Agentic AI and reinforcement learning. Following developments where NVIDIA and Oracle announced plans to scale Oracle Cloud Infrastructure Supercluster to a multi-gigawatt scale, NVIDIA has further expanded its Vera Rubin platform. At the SC26 Supercomputing conference, NVIDIA unveiled the Rubin HPC GPU, a chip designed for high-performance computing, scientific simulation, and engineering workloads. The Rubin HPC features more than 50,000 CUDA cores and utilizes seventh-generation NVLink interconnect technology alongside specialized sparse data acceleration to optimize tasks such as genomic analysis and climate modeling. NVIDIA expects the Rubin HPC to be available in 2028. NVIDIA also introduced the Vera CPU rack-scale platform, which integrates custom CPUs with Quantum-X800 InfiniBand networking to enhance throughput for molecular dynamics and weather simulations. This platform platform, which utilizes liquid cooling and integrates 192 Vera processors to achieve an aggregate peak memory bandwidth exceeding 200 TB/s, will be offered by partners including Dell Technologies, HPE, Lenovo, and Supermicro, aiming to provide higher energy efficiency and memory bandwidth for data-intensive scientific applications. Supermicro.
Versions
- 2026-11-18 15:22 UTC Supermicro and NVIDIA expand Vera Rubin AI infrastructure
- 2026-11-17 13:28 UTC Supermicro and NVIDIA expand Vera Rubin AI infrastructure
- 2026-11-16 20:16 UTC Supermicro and NVIDIA expand Vera Rubin AI infrastructure
- 2026-11-02 15:26 UTC Supermicro and NVIDIA expand Vera Rubin AI infrastructure
- 2026-10-25 06:46 UTC Supermicro ships NVIDIA Vera Rubin; unveils Agentic AI rack
- 2026-10-16 04:13 UTC Supermicro ships NVIDIA Vera Rubin; NVIDIA opens designs
- 2026-09-29 04:53 UTC Supermicro shipping NVIDIA Vera Rubin infrastructure
Select a version to see what changed compared to the one before it.