Supermicro Ships NVIDIA Blackwell Ultra Systems for AI Infrastructure

SAN JOSE, Calif., Sept. 11, 2025 /PRNewswire/ — Supermicro, Inc. (NASDAQ: SMCI) has announced the volume shipment of its NVIDIA Blackwell Ultra solutions, including NVIDIA HGX B300 systems and GB300 NVL72 racks, to customers worldwide.

The company’s Data Center Building Block Solutions (DCBBS) provide pre-validated and plug-and-play systems which enable rapid deployment at system, rack, and data center scales. These solutions are designed to support high-performance applications in artificial intelligence (AI), cloud services, and other data-intensive fields.

Key performance metrics for the GB300 NVL72 rack-scale system include 1.1 exaFLOPS of dense FP4 compute power and an impressive 7.5 times performance gain over previous NVIDIA Hopper systems. Each GPU in these systems boasts 144 petaFLOPS of FP4 compute capability and 270 GB of HBM3e memory.

Charles Liang, president and CEO of Supermicro, emphasized the company’s expertise in delivering advanced technology rapidly. “Our DCBBS enables turn-key delivery of the highest-performance AI platform, critical for customers,” he stated.

In addition to performance, the DCBBS implementation aims to optimize energy usage and resource conservation. Customers can expect up to 40% energy savings, a 60% reduction in data center footprint, and a 20% decrease in total cost of ownership (TCO).

Supermicro’s focus on creating efficient cooling solutions complements the high power demands of these systems, offering configurations that utilize up to 1400W per GPU while delivering advanced thermal management.

By effectively navigating supply chain challenges to secure these components, Supermicro positions itself favorably in the rapidly expanding AI infrastructure market, ready to capture significant market share amid ongoing upgrades in AI hardware.