NVIDIA Opens Vera Rubin Architecture at OCP to Accelerate AI Factory Scaling

As AI factories expand from megawatt to gigawatt scale, the industry faces a pressing need for interoperable, power-efficient infrastructure. At the Open Compute Project (OCP) Global Summit in San Jose, NVIDIA took a significant step by contributing its Vera Rubin rack and compute tray designs to the open MGX architecture, giving system builders a shared foundation for next-generation accelerated computing platforms.

Ecosystem Alignment on Open Standards

Major hyperscalers — including Cisco, Google, Meta, Microsoft, and Visionbay — rallied around the initiative. Together they published open AI factory design standards, grid resilience guidelines, and new networking specifications such as the Ultra-Dense Small Form Factor Pluggable (USFP) module for 400 Gb/s fabrics and Falcon Transport support on future ConnectX-10 SuperNICs and BlueField DPUs. Visionbay also unveiled an 800 V DC factory design tailored for the Vera Rubin era.

The updated MGX 1.2 specification extends the reference design with blind-mate liquid cooling at rack scale, a 5,000 A liquid-cooled busbar, native 19-inch equipment support, and configurable I/O bays. These enhancements streamline assembly, test, and service workflows, reducing deployment time for high-density AI clusters.

Why it matters for GPU / AI infrastructure: Open rack-level standards lower integration risk and accelerate time-to-revenue for cloud providers and enterprises building large-scale GPU clusters. By aligning power, cooling, and interconnect interfaces, the ecosystem can mix and match components from multiple vendors while maintaining the tokens-per-watt efficiency that defines modern AI factory economics.