Supermicro began shipping NVIDIA Vera Rubin NVL72 racks in configurations featuring 1,152 total accelerators across 16-rack blueprints, enabling large-scale training and inference.
Supermicro is now shipping NVIDIA Vera Rubin NVL72 racks integrated with the company's Data Center Building Block Solutions (DCBBS) and direct liquid cooling stack (DLC-2). The announcement underscores Supermicro's end-to-end capability in deploying large-scale AI infrastructure.
"We have spent years building the liquid-cooling stack, the manufacturing capacity, and the deployment teams for exactly this moment," said Charles Liang, president and CEO of Supermicro. "Our customers can now order a Scalable Unit and receive production-ready systems with end-to-end integration, because we design and build every layer between the cold plate and the cooling tower. Our development of DLC-2 technology is what sets Supermicro apart and allows us to deliver the most complete solution for the NVIDIA Vera Rubin platform."
The NVIDIA Vera Rubin platform was engineered specifically for direct liquid cooling to achieve AI throughput-per-watt efficiency levels unattainable with air cooling alone. Supermicro tests and validates every rack with the full liquid cooling stack—from cold plates through manifolds to cooling towers—to accelerate deployment timelines.
Each Vera Rubin NVL72 rack contains 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs operating as a single liquid-cooled machine. Eighteen 1U compute trays, each equipped with four Rubin GPUs and two Vera CPUs, connect through nine sixth-generation NVIDIA NVLink switch trays delivering 216 TB/s of scale-up bandwidth. Each rack provides 20.7 TB of HBM4 memory and up to 54 TB of LPDDR5X.
The integrated cooling solution includes Supermicro in-row cooling distribution units rated at 1.8 MW per CDU, deployed with N+1 redundancy, optional rear door heat exchangers for residual heat capture, and networking integration based on NVIDIA Reference Architecture. Supermicro's comprehensive liquid cooling portfolio—cold plates, manifolds, hose kits, rack power shelves, in-row and in-rack CDUs, sidecar CDUs, rear door heat exchangers, and cooling towers—is manufactured in-house.
Supermicro's DCBBS Blueprint simplifies deployment by defining a balanced bill-of-materials for a given power envelope, from 5 MW to gigawatt scale. A single Vera Rubin NVL72 Scalable Unit includes 1,152 NVIDIA Rubin GPUs with 331 TB of HBM4 memory distributed across 16 compute racks, with matched cooling capacity, power delivery, storage, context memory, and networking. A dedicated Supermicro team manages the entire engagement—site survey, design, integration, testing, delivery, deployment, and ongoing support.