Nvidia Open-Sources Vera Rubin Rack Designs via Open Compute Project
Nvidia has open-sourced its Vera Rubin data center rack designs through the Open Compute Project, contributing the specifications via its MGX 1.2 standard alongside industry partners including AMD, Google, Meta, and Cisco.
What the Open Compute Project Does
The Open Compute Project is an industry initiative that publishes open hardware designs for data center infrastructure, allowing manufacturers and cloud providers to build compatible equipment without licensing proprietary designs from a single vendor. By contributing Vera Rubin rack specifications through this channel, Nvidia is making it easier for data center operators and equipment makers to build infrastructure compatible with its next-generation AI hardware platform.
What’s in the Vera Rubin Designs
The contribution includes a new data center specification covering rack-level power, cooling, and networking requirements needed to support Nvidia’s upcoming Vera Rubin chip platform, the successor to its current Blackwell generation. Open rack standards help ensure that cooling systems, power delivery, and networking equipment from different manufacturers can work together reliably at the scale required for large AI training and inference clusters.
Why Nvidia Is Sharing the Design
Open-sourcing rack designs benefits Nvidia by encouraging broader industry adoption of infrastructure compatible with its chips, reducing friction for data center operators who might otherwise need custom engineering to deploy Nvidia’s latest hardware at scale. It also positions Nvidia’s MGX standard as a central reference point in the industry, similar to how earlier open rack standards shaped broader data center design practices.
Industry Collaboration Amid Competition
The involvement of AMD, a direct competitor to Nvidia in AI chips, alongside major cloud providers Google and Meta, reflects how infrastructure standards often see cooperation even among companies that compete fiercely on the chips themselves. Shared rack and power standards reduce costs and complexity across the industry, benefiting data center operators even as the underlying chip competition between Nvidia, AMD, and others remains intense. The move comes as AMD separately unveiled its own open Helion rack design, suggesting a broader industry push toward shared infrastructure standards even as chip-level competition intensifies.
