OCP Global Summit: Amid Ecosystem Adoption, NVIDIA Opens Vera Rubin Architecture to Help Scale AI Factories Worldwide

At the Open Compute Project Global Summit in San Jose, NVIDIA announced contributions to open standards, including Vera Rubin rack and compute tray designs, to create a common foundation for AI factory infrastructure. The company emphasized how open infrastructure accelerates deployment, improves power efficiency, and enhances tokens-per-watt performance in AI factories. NVIDIA’s contributions align with the third-generation MGX AI data center reference design, incorporating open liquid-cooling technologies like MQD and MQDB to turn proven designs into industry standards.

NVIDIA introduced the MGX 1.2 specification for Vera Rubin AI factories, featuring rack-scale blind-mate liquid cooling, interoperable interconnect infrastructure, and a liquid-cooled 5,000A busbar for higher-power systems. The specification also adds native 19-inch equipment support, higher-density racks, configurable I/O bays, and standardized power and cooling interfaces to streamline manufacturing and service times.

The Vera CPU rack design extends the NVIDIA MGX open architecture beyond accelerated computing, supporting agentic AI workloads such as tool calling and orchestration. Built on the third-generation MGX architecture, the Vera CPU rack shares modular design principles with compute, networking, power, and cooling to enable simplified integration and ecosystem reuse for next-generation AI factories.

NVIDIA also expanded its DSX OS with open source components for configuring networks, validating clusters, managing racks, and provisioning bare metal, derived from experience running NVIDIA DGX Cloud. The company joined industry leaders to publish the Open Data Center Specification v0.7 and the OCP Grid Resilience Specification, establishing unified blueprints for AI factory facility design and grid behavior during disturbances.