OCP Global Summit: NVIDIA Opens Vera Rubin Architecture to Help Scale AI Factories Worldwide

The core move NVIDIA announced at the OCP Global Summit was contributing key Vera Rubin rack and compute tray designs to OCP through the open NVIDIA MGX architecture, providing system builders a common foundation for developing next-generation AI factory infrastructure. The summit was held at the McEnery Convention Center in San Jose.

This contribution was framed within the narrative of "AI factories moving from megawatts to gigawatts": NVIDIA argues that building such facilities depends not only on advanced computing, and that open standards help the industry scale AI infrastructure faster, more efficiently and more reliably. In other words, open standards are positioned as a precondition for scale, not an option.

A distinction is needed: the contributed designs themselves are confirmed fact; while "open standards can accelerate deployment, improve energy efficiency, and boost tokens per watt performance" is NVIDIA's statement in its press release, and the source provides no independent verification data.

The updated MGX 1.2 specification for Vera Rubin AI factories expands the open AI factory architecture, adding rack-level blind-mate liquid cooling, interoperable rack-level interconnect infrastructure, and a liquid-cooled 5,000A busbar for higher-power AI systems.

The specification also adds native 19-inch device support and higher-density rack designs, and improves modularity through configurable I/O bays, fanless liquid-cooled compute trays, and standardized power and cooling interfaces. The architecture supports rack deployments up to 250kW and simplifies maintainability, deployment and infrastructure reuse.

These specifications all come from a unilateral NVIDIA release; the source shows no independent testing or certification, nor any actual deployment cases of 250kW racks. They should therefore be understood as specification targets and design capabilities, not third-party-verified operational data.