Supermicro expands enterprise AI portfolio to 100+ system SKUs supporting Nvidia Blackwell Ultra and HGX configurations.
Supermicro announced new systems and rack solutions powered by NVIDIA's Blackwell Ultra platform at GTC 2025, featuring the NVIDIA HGX B300 NVL16 and NVIDIA GB300 NVL72 platforms. The company, described as a Total IT Solution Provider for AI, Cloud, Storage, and 5G/Edge, stated that these solutions strengthen leadership in AI by delivering breakthrough performance for compute-intensive workloads including AI reasoning, agentic AI, and video inference applications.
Charles Liang, president and CEO of Supermicro, expressed enthusiasm for the partnership: "At Supermicro, we are excited to continue our long-standing partnership with NVIDIA to bring the latest AI technology to market with the NVIDIA Blackwell Ultra Platforms. Our Data Center Building Block Solutions® approach has streamlined the development of new air and liquid-cooled systems, optimized to the thermals and internal topology of the NVIDIA HGX B300 NVL16 and GB300 NVL72." He highlighted that the advanced liquid-cooling solution delivers exceptional thermal efficiency, operating with 40℃ warm water in the 8-node rack configuration, or 35℃ warm water in double-density 16-node rack configuration using their latest CDUs. Liang noted this innovative solution reduces power consumption by up to 40% while conserving water resources.
NVIDIA's Blackwell Ultra platform delivers an unprecedented 288GB HBM3e of memory per GPU and is engineered to overcome performance bottlenecks caused by limited GPU memory capacity and network bandwidth. The networking platform integration with NVIDIA Quantum-X800 InfiniBand and Spectrum-X™ Ethernet doubles the compute fabric bandwidth to up to 800 Gb/s. The platform delivers drastic improvements in AI FLOPS for both AI training and inference for the largest AI models.
Supermicro integrates NVIDIA Blackwell Ultra through two solution types: the NVIDIA HGX B300 NVL16 system designed for every data center, and the NVIDIA GB300 NVL72 equipped with NVIDIA's next-generation Grace Blackwell architecture. The new 8U HGX B300 NVL16 platform features 16 GPUs connected in a 1.8TB/s 16-GPU NVLink domain providing 2.3TB of HBM3e per system, with 8 NVIDIA ConnectX®-8 NICs integrated directly into the baseboard to support 800 Gb/s node-to-node speeds. The GB300 NVL72 integrates 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace™ CPUs in a single rack with exascale computing capacity, featuring over 20TB of HBM3e memory interconnected in a 1.8TB/s 72-GPU NVLink domain, with NVIDIA ConnectX®-8 SuperNIC providing 800Gb/s speeds for both GPU-to-NIC and NIC-to-network communication.
Supermicro's liquid cooling portfolio includes newly developed direct-to-chip cold plates, a 250kW in-rack CDU, and cooling tower. The company's on-site rack deployment services help enterprises build data centers from the ground up, including planning, design, power-up, validation, testing, installation and configuration of racks, servers, switches and other networking equipment to meet specific organizational needs.