Super Micro Computer Inc. (NASDAQ: SMCI) announced new server systems designed for NVIDIA's upcoming Vera Rubin platform, featuring liquid-cooling technology to support next-generation artificial intelligence workloads.

The company revealed three main system configurations: the NVIDIA Vera Rubin NVL72 SuperCluster, a 2U HGX Rubin NVL8 system, and NVIDIA Vera CPU systems. All systems utilize Supermicro's Data Center Building Block Solutions (DCBBS) liquid-cooling technology stack.

The Vera Rubin NVL72 operates as a single rack-scale accelerator that combines six co-designed chips including Rubin GPU, Vera CPU, NVIDIA NVLink 6, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet. According to the company, the system delivers up to 3.6 Exaflops of inference, 75TB of memory, and 1.6 PB/s of HBM4 bandwidth.

The 2U HGX Rubin NVL8 system supports eight Rubin GPUs and offers CPU flexibility, accommodating NVIDIA Vera CPUs alongside next-generation AMD and Intel x86 processors. The design allows for nine HGX Rubin NVL8 systems per rack, totaling up to 72 Rubin GPUs.

Supermicro's Vera CPU system features dual NVIDIA Vera CPUs supporting up to six RTX PRO 4500 Blackwell Server Edition GPUs in a 2U chassis. The company also introduced a Context Memory Storage Platform powered by NVIDIA BlueField-4 processor for extending GPU cache capacity.

"We are entering a new era where every organization requires an AI factory to win in the marketplace," said Charles Liang, president and CEO of Supermicro.

The DCBBS approach includes validated liquid-cooling infrastructure components such as coolant distribution units, manifolds, and liquid-to-air sidecars. The systems are designed to address the thermal and power requirements of liquid-cooled AI infrastructure.

Supermicro plans to showcase these systems at GTC San Jose 2026. The company's current NVIDIA Blackwell-based systems remain in production and available for immediate deployment.