Supermicro Now Shipping NVIDIA Vera Rubin NVL72 Racks

Super Micro Computer announced that it has begun shipping NVIDIA Vera Rubin NVL72 racks integrated with Supermicro's Data Center Building Block Solutions (DCBBS) and its direct liquid cooling stack (DLC-2). The enterprise deployment solution is designed to accelerate time-to-online for large-scale AI data center infrastructure, supporting deployments across power envelopes ranging from 5 megawatts up to gigawatt scale. Each NVIDIA Vera Rubin NVL72 rack integrates 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs operating as a single coherent machine. The system delivers 216 TB/s of scale-up bandwidth via sixth-generation NVIDIA NVLink switch trays, paired with 20.7 TB of HBM4 memory and up to 54 TB of LPDDR5X. A complete DCBBS Scalable Unit blueprint comprises 16 compute racks incorporating 1,152 NVIDIA Rubin GPUs and 331 TB of HBM4 memory. The turnkey liquid cooling stack spans cold plates, in-row cooling distribution units rated at 1.8MW per CDU with N+1 redundancy, and cooling towers to manage the thermal dissipation demanded by high-density AI clusters.