THE CRUNCH
NVIDIA has opened up its Vera Rubin rack and compute tray designs to the Open Compute Project through its MGX architecture, giving system builders a shared blueprint for next-generation AI factory infrastructure. The announcement came at the OCP Global Summit in San Jose, where NVIDIA also published open standards with hyperscalers and released new open source software for its DSX AI Factory platform.
The updated MGX 1.2 specification extends the open architecture with rack-scale blind-mate liquid cooling, a liquid-cooled 5,000A busbar for higher-power systems, native 19-inch equipment support and denser rack designs. It supports rack deployments up to 250kW while aiming to shorten assembly, testing and service times. A separate Vera CPU rack design targets agentic AI workloads such as tool calling, orchestration and sandboxed execution.
The software side adds four open source DSX OS components for configuring networks, stress-testing GPU clusters before training, managing racks as single units and provisioning bare metal, the last of which NVIDIA says cuts cluster setup from days to hours. Partner announcements included an 800V DC AI factory design from Visionbay, a USFP standard with Cisco for 400Gb/s networking modules, and Falcon Transport support on future ConnectX-10 SuperNICs and BlueField DPUs with Google.