By Harshita Tyagi
Publication Date: 2026-09-23 14:00:00
Supermicro is moving Nvidia’s latest AI platform into customer deployments while bundling the servers with cooling, power and other data-center infrastructure.
- Each NVL72 rack combines 72 Rubin GPUs and 36 Vera CPUs.
- Supermicro’s 16-rack design scales to 1,152 GPUs and 331 TB of HBM4 memory.
- The systems are part of Supermicro’s broader push to supply complete AI data-center infrastructure, not just servers.
Super Micro Computer Inc. (SMCI) said it has begun shipping Nvidia Corp. (NVDA) Vera Rubin NVL72 racks, bringing Nvidia’s latest AI computing platform into production deployments alongside Supermicro’s liquid cooling and data center infrastructure.
The systems are being offered through Supermicro’s Data Center Building Block Solutions, which combine compute, cooling, networking, storage and power infrastructure to help customers bring large AI clusters online more quickly.
What Is Inside SMCI’s NVL72?
Each rack contains 72 Nvidia Rubin GPUs and 36 Vera CPUs, connected to work together as a single large AI computing system.
It also includes 20.7 TB of HBM4, or high-bandwidth memory, which, according to Nvidia, is designed to move large amounts of data quickly between processors and support increasingly demanding AI workloads.
Nvidia has positioned Vera Rubin for agentic AI, where AI systems can reason through multiple steps, use tools, and continue working on tasks rather than simply answering individual prompts.
Supermicro Scales Up To 1,152 GPUs
SMCI’s larger…