Nvidia shares Vera Rubin rack and tray designs with OCP

Nvidia said on October 12 that it had contributed Vera Rubin rack and compute-tray designs to the Open Compute Project (OCP). Announced at the OCP Global Summit, the contribution uses the company's open MGX architecture to give system builders a common starting point for AI infrastructure.

According to Nvidia, MGX 1.2 supports rack deployments of up to 250 kilowatts. The specification includes blind-mate liquid cooling, interoperable rack interconnects and a liquid-cooled busbar specified at 5,000 amps. It also accommodates 19-inch equipment and adds configurable I/O bays.

The update includes fanless compute trays that use liquid cooling, along with standardized power and cooling interfaces. Nvidia says the architecture reduces time spent assembling, testing and servicing systems. Those time-saving claims are not accompanied by independent measurements in the two published accounts.

A separate Vera CPU rack specification extends MGX to a platform aimed at agent tool calling, orchestration and sandboxed execution. Data Center Dynamics reports that Nvidia says the CPU rack has started shipping. That report concerns the CPU rack specifically, rather than establishing availability for all Vera Rubin infrastructure.

Analysis: shared rack, power and cooling specifications could help suppliers prepare compatible components without waiting for each customer's proprietary requirements. The potential benefit is less integration work for infrastructure builders; the announcement itself does not establish the scale of any resulting cost savings.