Shares of Super Micro Computer (NASDAQ:SMCI) rose 8% on Monday following the company's announcement at ISC 2026 of a Data Center Building Block Solutions Blueprint tailored to high-performance computing on the NVIDIA Vera Rubin NVL4 platform.
The blueprint describes an integrated, liquid-cooled architecture that can house as many as 1,152 NVIDIA Rubin GPUs and 576 NVIDIA Vera CPUs in a single 3.2MW Scalable Unit. Supermicro positions the design to handle demanding AI workloads alongside FP64 double-precision simulation tasks.
At the hardware level, the solution is composed of eight liquid-cooled compute racks fitted into customized 52U, 750mm-wide enclosures. Each rack contains 36 NVIDIA Vera Rubin NVL4 nodes within a 362 kW electrical envelope, yielding 288 nodes across the Scalable Unit. Supermicro’s DLC-2 Direct Liquid Cooling technology is specified to support 362 kW per rack, with three in-row cooling distribution units allocated to each Scalable Unit.
The announcement highlights the target applications for the NVIDIA Vera Rubin NVL4 platform, emphasizing scientific computing workloads that blend traditional FP64 simulation with AI methods. Fields called out for this configuration include climate research, drug discovery, materials science, and energy.
Beyond the hardware specification, Supermicro’s blueprint is described as an end-to-end deployment package. The company says the offering begins with on-site facility surveys to evaluate factors such as loading dock access and power infrastructure, continues through manufacturing and testing at Supermicro’s facilities, and concludes with on-site integration tasks including rack placement and network cabling. The blueprint also includes ongoing support commitments, with on-site response times described as fast as four hours.
This NVL4 blueprint follows earlier Data Center Building Block Solutions (DCBBS) Blueprints that the company introduced at Computex for the NVIDIA Vera Rubin NVL72 and the NVIDIA HGX Rubin NVL8 platforms. Supermicro also noted that configurations based on the NVIDIA GB200 NVL4 are available for immediate deployment.
The company’s description focuses on delivering a packaged pathway for organizations seeking scalable, liquid-cooled data center capacity designed for mixed AI and high-precision simulation workloads, while accounting for the facility-level requirements and on-site integration necessary to bring such systems online.