At the October 12 OCP Global Summit, NVIDIA shared Vera Rubin rack designs and announced infrastructure standards and partner developments for AI factories.
NVIDIA announced new open infrastructure contributions at the Open Compute Project (OCP) Global Summit on October 12, 2026, including rack and compute-tray designs for its Vera Rubin systems. The designs were contributed through NVIDIA’s MGX architecture, giving equipment makers and infrastructure operators a shared reference for building AI data centers.
The updated MGX 1.2 specification supports rack deployments of up to 250 kilowatts. It includes rack-scale liquid cooling, interoperable interconnects, a liquid-cooled 5,000-amp busbar, native 19-inch equipment support and higher-density rack configurations. NVIDIA also described a Vera CPU rack design intended for tasks such as tool calling, orchestration and sandboxed execution, alongside the Vera Rubin NVL72 platform for large-scale training and inference.
The announcements extended beyond rack hardware. NVIDIA joined AMD, Google and Meta in publishing Open Data Center Specification v0.7, a blueprint covering facility design elements including power, cooling and physical layout. Google, Meta, Microsoft and NVIDIA also published an OCP Grid Resilience Specification describing how large AI facilities should respond to utility disruptions, including a return to near-full load in less than two seconds.
Other partner announcements covered networking and operations. Meta plans to use NVIDIA Spectrum-6 switches in its AI infrastructure, while NVIDIA and Cisco announced the USFP optical-transceiver form factor, designed for 400-gigabit-per-second-per-lane networking. NVIDIA also introduced open-source DSX software components for network configuration, cluster checks, rack management and bare-metal provisioning.
Together, the contributions address several layers of AI data-center construction, from facility standards and power resilience to racks, networking and software operations. The specifications and designs are intended to give infrastructure suppliers and operators common foundations as they develop systems for Vera Rubin and other large-scale AI deployments.