Nvidia announced at the OCP Global Summit 2026 the contribution of Vera Rubin platform rack and tray designs to the Open Compute Project (OCP) through the open MGX architecture. The move aims to accelerate the construction of AI factories on a global scale.
What happened?
During the event in San Jose, the company contributed MGX 1.2 specifications, including rack-scale blind-mate liquid cooling, interoperable interconnect infrastructure and a liquid-cooled 5,000A busbar. The designs support racks of up to 250 kW, with compute trays without cables, hoses or fans.
Joe DeLaere of Nvidia highlighted that the current bottleneck is the speed of manufacturing and deployment. Opening the specifications allows manufacturers to start producing immediately, expanding the ecosystem.
Why does this matter?
The Vera Rubin platform is designed for the agentic AI era, with NVL72 racks and specialized systems for orchestration. By standardizing via OCP, Nvidia facilitates adoption by partners such as Meta, Google, Cisco and Microsoft, who are already integrating related technologies such as Spectrum-6.
This helps scale from megawatts to gigawatts more efficiently, with better cooling and modularity.
What changes in practice?
Manufacturers and data center operators can build compatible infrastructure faster, reducing costs and deployment time. The open approach accelerates the growth of the AI hardware market without depending on exclusive proprietary designs.
Sources: NVIDIA Blog, SDxCentral.
Image credit: NVIDIA — Source: official announcement
By GeekikiBot
