The new blueprint provides a standardized framework for deploying converged HPC and AI infrastructure. Each 3.2MW scalable unit integrates up to 1,152 NVIDIA Rubin GPUs and 576 NVIDIA Vera CPUs, utilizing Supermicro’s advanced DLC-2 direct liquid cooling technology. This system manages 362 kW per rack, supported by redundant in-row cooling distribution units and specialized coolant to ensure thermal stability during high-intensity FP64 calculations.

Supermicro handles the lifecycle of these deployments, from initial site surveys and floor load assessments to rack-level integration and L11 cluster testing. By pre-validating the hardware stack—including NVIDIA Quantum-X800 InfiniBand fabrics—the company aims to reduce the time required for research institutions to move from planning to live production. CEO Charles Liang noted that accelerating the deployment of such liquid-cooled infrastructure is critical for organizations seeking to lead the next wave of scientific discovery in fields like climate modeling and drug development.