Nvidia has contributed Vera Rubin rack and compute tray designs to the Open Compute Project. Its MGX 1.2 specification connects liquid cooling, power and rack infrastructure through common interfaces. The announcement also includes released management software for AI facilities, giving manufacturers and operators design documents and tools rather than reporting completed construction at every site.
Nvidia announced on October 12 that it had contributed Vera Rubin rack and compute tray designs to the Open Compute Project, an industry initiative for shared hardware designs. The contribution uses MGX, Nvidia’s modular infrastructure architecture. System builders receive a common foundation covering the computing equipment and its power, cooling and connections.[1], [2]
MGX 1.2 defines shared interfaces for rack-scale liquid cooling and interconnect infrastructure. It includes fanless, liquid-cooled compute trays, configurable input-output bays and a cooled power busbar for high-power systems. Nvidia describes easier manufacturing, testing and maintenance as design goals. The Vera CPU rack design also addresses AI agents’ tool calls, task coordination and code execution in isolated environments.[1]
Nvidia says it released open-source components for its DSX facility-management platform, drawing on its operation of DGX Cloud. They cover switch configuration, cluster-readiness tests, rack hardware management and installation directly on bare hardware. Further open facility specifications cover power, cooling, physical layout and responses to electricity-grid disturbances. These are published designs and software components; adoption and facility construction remain separate stages.[1]
