NVIDIA Opens Vera Rubin Architecture at OCP Summit to Scale AI Factories

At the Open Compute Project (OCP) Global Summit at the San Jose McEnery Convention Center, NVIDIA announced new partnerships and contributions aimed at creating a common foundation for next-generation AI infrastructure. As AI factories grow from megawatts to gigawatts, the company argues that advanced computing alone is not enough and that open standards are what let the industry scale infrastructure faster, more efficiently and more reliably.

The centerpiece is NVIDIA's decision to contribute key Vera Rubin rack and compute tray designs to OCP through its open MGX architecture, giving system builders a common starting point for next-generation AI factory infrastructure. Together, the contributions span the AI factory itself, the utility grid that powers it and the rack-scale architecture at the heart of modern accelerated computing.

Alongside its own work, NVIDIA and partners made a handful of announcements: joint open standards for next-generation AI factories covering design specifications and grid resilience; new open source software components for the NVIDIA DSX AI Factory platform; and, with Meta, advances in open AI networking built on the NVIDIA Spectrum-X Ethernet SuperNIC and Spectrum-6.

The announcements also included an 800V DC AI factory design for the Vera Rubin era from Visionbay; an Ultra-Dense Small Form Factor Pluggable (USFP) standard from NVIDIA and Cisco, which they describe as the industry's first module for 400Gb/s-based networking fabrics; and Falcon Transport support on future NVIDIA ConnectX-10 SuperNICs and BlueField DPUs with Google.

In the updated MGX 1.2 specification, the open architecture for Vera Rubin AI factories expands further with rack-scale blind-mate liquid cooling, interoperable rack-scale interconnect infrastructure and a liquid-cooled 5,000A busbar designed for higher-power AI systems. It also adds native 19-inch equipment support and higher-density rack designs.

The architecture supports rack deployments up to 250kW and adds configurable I/O bays, fanless liquid-cooled compute trays and standardized power and cooling interfaces to improve modularity, simplify assembly, test and service, and speed up infrastructure buildout. Beyond the Vera Rubin NVL72 system for large-scale training and inference, NVIDIA also detailed a Vera CPU rack design aimed at tool calling, orchestration and sandboxed execution for agentic AI.

On the software side, NVIDIA expanded DSX OS, the open operating layer of its DSX AI factory platform, with open source components for configuring networks, validating clusters, managing racks and provisioning bare metal, so that an AI factory can be run as a single coordinated, repeatable system.

For the industry, the significance lies in standardizing the heavy, capital-intensive layers of AI factories, from liquid cooling to power and rack interconnect. If these designs are widely adopted, deployment and reuse costs should fall; whether the ecosystem genuinely adopts them, and whether interoperability between vendors holds up, is what the next few quarters will test.