Supermicro Ships New Liquid-Cooled Systems for Data Centers

Super Micro Computer has expanded its NVIDIA Blackwell architecture portfolio with the introduction and shipment availability of new 4U and 2-OU (OCP) liquid-cooled NVIDIA HGX B300 systems. These latest additions are a part of Supermicro's Data Center Building Block Solutions (DCBBS) that deliver "unprecedented GPU density and power efficiency for hyperscale data centers and AI factory deployments," the company said.

"With AI infrastructure demand accelerating globally, our new liquid-cooled NVIDIA HGX B300 systems deliver the performance density and energy efficiency that hyperscalers and AI factories need today," said Charles Liang, president and CEO of Supermicro. "We're now offering the industry's most compact NVIDIA HGX B300 solutions—achieving up to 144 GPUs in a single rack—while reducing power consumption and cooling costs through our proven direct liquid-cooling technology. Through our DCBBS, this is how Supermicro enables our customers to deploy AI at scale: faster time-to-market, maximum performance per watt, and end-to-end integration from design to deployment."

The 2-OU (OCP) liquid-cooled NVIDIA HGX B300 system, built to the 21-inch OCP Open Rack V3 (ORV3) specification, enables up to 144 GPUs per rack to deliver maximum GPU density for hyperscale and cloud providers requiring space-efficient racks without compromising serviceability. The rack-scale design features blind-mate manifold connections, modular GPU/CPU tray architecture, and state-of-the-art component liquid cooling solutions. The system propels AI workloads with eight NVIDIA Blackwell Ultra GPUs at up to 1,100W TDP each, while dramatically reducing rack footprint and power consumption. A single ORV3 rack supports up to 18 nodes with 144 GPUs total, scaling seamlessly with NVIDIA Quantum-X800 InfiniBand switches and Supermicro's 1.8MW in-row coolant distribution units (CDUs). Combined, eight NVIDIA HGX B300 compute racks, three NVIDIA Quantum-X800 InfiniBand networking racks, and two Supermicro in-row CDUs form a SuperCluster scalable unit with 1,152 GPUs.

Complementing the 2-OU (OCP) model, the 4U Front I/O HGX B300 Liquid-Cooled System offers the same compute performance in a traditional 19-inch EIA rack form factor for large-scale AI factory deployments. The 4U system leverages Supermicro's DLC-2 technology to capture up to 98% of heat generated by the system through liquid-cooling, achieving superior power efficiency with lower noise and greater serviceability for dense training and inference clusters.

New Supermicro NVIDIA HGX B300 systems. Image courtesy of Supermicro.

Supermicro NVIDIA HGX B300 systems unlock substantial performance speedups, with 2.1TB of HBM3e GPU memory per system to handle larger model sizes at the system level. Above all, both the 2-OU (OCP) and 4U platforms deliver significant performance gains at the cluster level by doubling compute fabric network throughput up to 800Gb/s via integrated NVIDIA ConnectX-8 SuperNICs when used with NVIDIA Quantum-X800 InfiniBand or NVIDIA Spectrum-4 Ethernet. These improvements accelerate heavy AI workloads such as agentic AI applications, foundation model training, and multimodal large scale inference in AI factories.

Supermicro developed these platforms to address key customer requirements for TCO, serviceability, and efficiency. With the DLC-2 technology stack, data centers can achieve up to 40 percent power savings, reduce water consumption through 45°C warm water operation and eliminate chilled water and compressors in data centers. Supermicro DCBBS delivers the new systems as fully validated, tested racks ready as L11 and L12 solutions before shipment, accelerating time-to-online for hyperscale, enterprise, and federal customers.

These new systems expand Supermicro's broad portfolio of NVIDIA Blackwell platforms — including the NVIDIA GB300 NVL72, NVIDIA HGX B200, and NVIDIA RTX PRO 6000 Blackwell Server Edition. Each of these NVIDIA-Certified Systems from Supermicro are tested to validate optimal performance for a wide range of AI applications and use cases – together with NVIDIA networking and NVIDIA AI software, including NVIDIA AI Enterprise and NVIDIA Run:ai. This provides customers with flexibility to build AI infrastructure that scales from a single node to full-stack AI factories.

Sources: Press materials received from the company and additional information gleaned from the company’s website.