US Server Maker Supermicro Announces NVIDIA Vera CPU Rack for Large-Scale Agentic AI
Supermicro launches a liquid-cooled rack featuring 256 NVIDIA Vera CPUs designed specifically for agentic AI workloads.
US server maker Super Micro Computer, Inc. announced on October 27, 2026, the latest addition to its Data Center Building Block Solutions (DCBBS) portfolio: the NVIDIA Vera CPU Rack. The product provides high-density CPU computing capabilities tailored for agentic AI, reinforcement learning workloads, and data processing.
The rack integrates 256 NVIDIA Vera CPUs, delivering a total of 22,528 cores and 45,056 threads, and supports up to 22,500 concurrent agent sandboxes in a liquid-cooled AI factory environment. It features a 100% liquid-cooled design based on the third-generation NVIDIA MGX architecture, allowing it to operate alongside or independently from Supermicro's currently shipping NVIDIA Vera Rubin NVL72 systems.
Charles Liang, President and CEO of Supermicro, stated that as agentic AI changes data center requirements, the company is expanding its DCBBS portfolio with fully integrated infrastructure. Chris Marriott, Vice President of Enterprise Platforms at NVIDIA, noted that enterprises scaling agentic AI and reinforcement learning require a new class of dedicated high-density CPU infrastructure to support thousands of simultaneous workflows.
Additionally, the rack integrates the NVIDIA BlueField-4 DPU for high-performance networking and chip-level security, and offers an 800V DC power option to improve space utilization and power conversion efficiency. The product is currently available for quote.