NVIDIA contributed Vera Rubin rack and compute tray designs to the Open Compute Project (OCP) through its open NVIDIA MGX architecture. At the OCP Global Summit in San Jose, NVIDIA and partners also published open standards for AI factory design and grid resilience, so system builders can work from a common target.

Key Features at a Glance

• MGX 1.2 specification: Adds rack-scale blind-mate liquid cooling, a liquid-cooled 5,000A busbar and native 19-inch equipment support, with rack deployments up to 250kW. This gives builders a shared design for higher-power Vera Rubin systems.
• Vera CPU rack design: Extends MGX to a rack for tool calling, orchestration and sandboxed execution. It is meant for AI agents that interact with applications, data and services in real time, alongside Vera Rubin NVL72 for training and inference.
• Spectrum-6 switches: Meta is adopting NVIDIA Spectrum-6 102.4Tb/s Ethernet switches. Spectrum-6 delivers twice the switching capacity and per-lane bandwidth of Spectrum-4, with a lossless Ethernet AI fabric at 200G per lane.
• Open Data Center Specification v0.7: Published with AMD, Google and Meta, it defines a unified blueprint for AI factory facility design, covering power architecture, cooling infrastructure and physical layout.
• OCP Grid Resilience Specification: Published by Google, Meta, Microsoft and NVIDIA, it defines how AI infrastructure should behave during utility grid disturbances: staying connected, responding automatically and restoring to near-full load within less than two seconds.
• DSX OS open source components: NVIDIA Config Manager automates switch configuration with minimal to zero manual intervention. NVIDIA Cluster Readiness Engine stress-tests GPU clusters with real workload simulations before training begins. Developed from NVIDIA's experience running DGX Cloud, these components help cloud providers integrate AI factory capabilities into multi-tenant platforms.

Visionbay also unveiled K-1, an AI factory platform built on an 800 VDC architecture with NVIDIA Vera Rubin in mind. It combines NVIDIA's 800 VDC reference architecture with power and thermal technologies from Vertiv. Rather than requiring a full transition on day one, K-1 phases in 800 VDC distribution where it delivers the most benefit, using cluster-level power centers that centralize AC-to-DC conversion.

Availability

The announcements were made at the OCP Global Summit in San Jose; the article gives no pricing or specific shipping dates for these designs and specifications.