Super Micro Computer, Inc. (hereinafter referred to as Supermicro) announced a structural change in government AI infrastructure architecture through a strategic collaboration with NVIDIA to deliver next-generation AI infrastructure.

Supermicro today announced its next-generation AI infrastructure solutions based on the NVIDIA Vera Rubin NVL144 and NVL144 CPX architectures. These new products are high-performance AI platforms designed to meet the security, performance, and regulatory requirements of the U.S. federal government.

The Vera Rubin platform delivers AI attention acceleration performance that is more than three times better than its predecessor, the Blackwell Ultra.

This enables efficient processing of next-generation AI applications such as large-scale language model (LLM) training, complex inference, and multimodal AI workloads. Based on this, Supermicro plans to provide infrastructure optimized for various fields such as AI/ML, HPC, cloud, storage, and 5G/edge.

“With our unique ability to conduct R&D and manufacturing right here in Silicon Valley, Supermicro is positioning itself as the federal government’s trusted AI infrastructure partner,” said Charles Liang, CEO of Supermicro.

Supermicro develops, validates, and manufactures all government systems at its San Jose, California headquarters. This ensures compliance with the Trade Agreements Act (TAA) and the Buy American Act (BAA), ensuring supply chain security and quality assurance.

Additionally, Supermicro supports the NVIDIA AI Factory for Government reference design. This architecture is designed to securely operate multiple AI workloads across on-premises and hybrid cloud environments, achieving both security and scalability.

Supermicro supports a variety of federal government workloads, including cybersecurity and threat detection, healthcare and life sciences research, data analytics and fusion platforms, and modeling, simulation, and secure virtual infrastructure, with its latest solutions based on NVIDIA HGX B300/B200, GB300/GB200, and RTX PRO 6000 Blackwell server GPUs.

In particular, the newly released 2OU HGX B300 8-GPU solution applies Supermicro's Data Center Building Block Solution (DCBBS)-based OCP rack-scale architecture to support up to 144 GPUs per rack.

It provides a high-density, high-performance computing environment required for large-scale AI and HPC deployment.

Additionally, by adding the NVIDIA GB300-based Super AI Station and GB200 NVL4 HPC solution, we are responding to the government agencies' demands for next-generation AI infrastructure.