started · updated
NVIDIA opens Vera Rubin AI platform designs to standardize data centers
NVIDIA is accelerating the standardization of AI data center infrastructure by opening its next-generation Vera Rubin AI platform rack and compute tray designs to the industry. Announced at the Open Compute Project (OCP) Global Summit, this move utilizes the third-generation MGX open architecture to allow system manufacturers to develop, manufacture, and expand AI racks using common specifications. This initiative aims to reduce integration complexity and speed up the deployment of global AI factories.
The updated MGX 1.2 specification supports rack deployments of up to 250kW and introduces features such as blind-mate liquid cooling, interoperable rack-level interconnects, and liquid-cooled power busbars with rated currents of 5,000 amps. The modular design is intended to simplify assembly, testing, and maintenance while improving compatibility between different suppliers.
In support of this ecosystem, Lite-On showcased its power solutions at the OCP summit, including 800 VDC power racks and 800V-to-50V power bricks. Lite-On’s technology is designed to meet the high power density and megawatt-scale computing demands of modern AI data centers, offering configurations ranging from 330 kW to 1.2 MW to support the transition toward high-power AI infrastructure.
Entities
Lite-On · MGX · Nvidia · Open Compute Project · Vera Rubin