Summarize with AI
Nvidia has contributed its Vera Rubin rack and compute tray designs to the Open Compute Project (OCP), so any hardware maker can build compatible systems. It also helped publish an open data center standard with AMD, Google and Meta, and released open-source tools for setting up AI clusters.
Nvidia announced this on October 12, 2026, at the OCP Global Summit in San Jose. The designs go out through MGX, Nvidia's open modular architecture for servers and racks.
What it is
- Vera Rubin racks and compute trays. Built on the updated MGX 1.2 specification, part of the third-generation MGX reference design.
- A Vera CPU rack. A rack design for agentic AI on the same third-generation MGX base. Nvidia says it handles tool calling, orchestration and sandboxed execution.
- Open Data Center Specification v0.7. Published by Nvidia, AMD, Google and Meta. It covers facility design: power, cooling and physical layout.
- DSX OS components, open source. Tools for preparing a cluster before it runs work (listed below).
MGX 1.2 at a glance
| Feature | Value |
|---|---|
| Rack power | up to 250 kW |
| Busbar | liquid-cooled, 5,000 A |
| Cooling | rack-scale blind-mate liquid cooling, fanless liquid-cooled compute trays |
| Equipment | native 19-inch support, configurable I/O bays |
The open-source DSX OS tools
| Tool | What it does |
|---|---|
| Config Manager | Configures network switches |
| Cluster Readiness Engine | Stress-tests GPU clusters |
| Rack Management Service | Manages rack firmware |
| Cluster API Provider for NVIDIA Infrastructure Controller | Provisions bare-metal servers |
Google, Meta, Microsoft and Nvidia also published an OCP Grid Resilience Specification. It says systems should stay connected during grid disturbances and get back to near-full load in under two seconds.
Sources
Did this land? Leave a reaction.
Discussion
0 commentsMembers can join the discussion
Become a member to comment and reply to other members.
Become a MemberNo comments yet
Be the first to start the discussion.