Supermicro launches AI Factory cluster solutions leveraging Nvidia's Blackwell GPU architecture for large-scale AI deployment.

Supermicro announced new complete, turnkey AI factory solutions at the Supercomputing Conference, based on NVIDIA Enterprise Reference Architectures and NVIDIA Blackwell GPUs. The rack-scale clusters are fully integrated and validated by Supermicro and delivered as full-stack solutions, complete with the NVIDIA software stack and NVIDIA Spectrum-X Ethernet networking. Charles Liang, president and CEO of Supermicro, stated: "Supermicro has always led the industry in time-to-market for new GPU technologies at rack-scale, and we're leveraging our expertise in delivering large-scale NVIDIA GB300 NVL72 and NVIDIA HGX B300-based AI infrastructure to enable the democratization of AI for enterprises in every industry." He further noted that "The AI factory is the foundation for transforming every company into an AI company, and in combination with our Data Center Building Block Solutions, Supermicro and NVIDIA are helping enterprises accelerate and streamline the deployment of AI factories for the industry's shortest time-to-online (TTO)." Supermicro's DCBBS can facilitate the build-out of AI factories, providing everything needed to develop a new greenfield data center or to refurbish a traditional data center into an AI factory.

Supermicro's AI Factory solutions are available in small, medium, and large configurations, ranging from 4 nodes and 32 GPUs up to 32 nodes and 256 GPUs. The clusters are integrated and tested up to L12 (multi-rack cluster) at Supermicro's global production sites, including in San Jose, California, and include the NVIDIA software stack (NVIDIA AI Enterprise, NVIDIA Omniverse, NVIDIA Run:ai), NVIDIA Spectrum-X Ethernet networking, and fully integrated cabling for a complete solution that is plug-and-play and ready to start generating tokens from day one. The solutions are optimized for environments where power, cooling, and physical space are common restricting factors when deploying AI infrastructure. Supermicro is currently taking orders for AI factory cluster solutions in 4, 8, and 32-node configurations with NVIDIA RTX PRO 6000 Blackwell Server Edition or NVIDIA HGX B200 GPUs, optimized for AI and HPC workloads at any scale.

Supermicro offers two specific cluster types. Universal AI, HPC, and visual computing clusters are based on 4U and 5U PCIe GPU systems with 8 GPUs per node in a 2-8-5-200 (CPU-GPU-NIC-Bandwidth) configuration, enabling enterprises to handle AI inference, enterprise HPC, and graphics and rendering workloads while leveraging common infrastructure for multiple applications. High performance AI and HPC clusters are based on 10U modular GPU platforms, with each node including NVIDIA HGX B200 8-GPU and supporting NVIDIA NVLink for maximum GPU-GPU communication, optimized for AI model fine-tuning and training, as well as HPC workloads. Supermicro also offers storage solutions for all stages of the AI data pipeline and can support the NVIDIA AI Data Platform reference design to enable AI workflows within the AI factory.