At the Open Compute Project Global Summit, NVIDIA announced it contributed Vera Rubin rack and compute tray designs to OCP through the open MGX architecture. The company says this gives system builders a common foundation for building AI factory infrastructure, spanning compute, power, cooling, and networking.
The updated MGX 1.2 specification extends the open design with rack-scale blind-mate liquid cooling, interoperable interconnect infrastructure, and a liquid-cooled 5,000A busbar. It adds native 19-inch equipment support and supports rack deployments up to 250kW, while simplifying serviceability and infrastructure reuse.
NVIDIA also released open source DSX OS components, including Config Manager, Cluster Readiness Engine, Rack Management Service, and a Cluster API provider for its infrastructure controller. These are designed to reduce manual intervention in network configuration, stress-test GPU clusters before training, and turn raw hardware into production-ready clusters in hours.
The Vera CPU rack design, also based on third-generation MGX, targets the agentic AI era with support for tool calling, orchestration, and sandboxed execution, complementing Vera Rubin NVL72 for large-scale training. As the article is based on a single NVIDIA announcement, no independent or contrasting perspectives are reported.