Supermicro is deepening its collaboration with NVIDIA to deliver new AI infrastructure platforms tailored for U.S. federal government needs, emphasizing secure, domestically manufactured systems. Announced at NVIDIA GTC in Washington, D.C., the company will bring the NVIDIA Vera Rubin NVL144 and CPX systems to market in 2026, providing over three times the AI attention acceleration of NVIDIA’s Blackwell Ultra. All systems will be manufactured and validated in San Jose, California, ensuring full compliance with the Trade Agreements Act (TAA) and Buy American Act requirements.
The expanded portfolio includes Supermicro’s compact 2OU NVIDIA HGX B300 8-GPU server with OCP-based rack-scale design supporting up to 144 GPUs in a single rack. Supermicro also unveiled the liquid-cooled Super AI Station, a deskside AI supercomputer based on the NVIDIA GB300 Superchip, capable of 20 PFLOPS of AI performance and supporting trillion-parameter models. Additionally, the company introduced its GB200 NVL4 HPC rack-scale system featuring four NVIDIA Blackwell GPUs and two Grace CPUs per node, enabling up to 128 GPUs in a 48U rack for advanced scientific and defense workloads.
Supermicro confirmed readiness to integrate NVIDIA’s new BlueField-4 DPUs and ConnectX-9 SuperNICs to accelerate networking, storage, and data offload in gigascale AI factories. These systems will support the NVIDIA AI Factory for Government reference architecture—an end-to-end framework designed for secure, on-premises AI model training and inference while meeting compliance standards for high-assurance environments.
• Vera Rubin NVL144 and CPX platforms deliver 3x acceleration vs. Blackwell Ultra
• 2OU HGX B300 8-GPU system supports 144 GPUs per rack with OCP-based design
• Super AI Station with NVIDIA GB300 offers up to 20 PFLOPS AI in a deskside form factor
• GB200 NVL4 rack-scale platform scales to 128 GPUs and 800G networking per GPU
• U.S.-based manufacturing in San Jose ensures TAA and Buy American compliance
“Our expanded collaboration with NVIDIA and our focus on U.S.-based manufacturing position Supermicro as a trusted partner for federal AI deployments,” said Charles Liang, president and CEO of Supermicro.
🌐 Analysis:
This announcement reinforces Supermicro’s strategic position as a core hardware partner in NVIDIA’s expanding AI ecosystem—particularly in secure, domestically manufactured systems for federal agencies. The company’s focus on compliance and liquid-cooled rack-scale designs parallels broader moves by U.S. hyperscalers and government programs to localize AI compute infrastructure. NVIDIA’s Vera Rubin and BlueField-4 platforms, paired with Supermicro’s modular Building Block architecture, signal a continued push toward scalable, secure, and power-efficient AI infrastructure for mission-critical applications.
