The development

What happened

NVIDIA is contributing Vera Rubin rack and compute-tray reference designs through MGX to the Open Compute Project, alongside open-source tools for operating AI infrastructure. Announced on October 12, the work addresses the surrounding power, cooling, networking and assembly systems needed to build larger AI datacenters. The contribution concerns shared infrastructure designs and software; it does not open-source Rubin GPU silicon.

Key facts

  • MGX 1.2 adds blind-mate liquid cooling, common power and cooling interfaces, configurable I/O and support for standard 19-inch equipment.

  • NVIDIA describes a liquid-cooled 5,000A busbar and rack configurations supporting up to 250 kW.

  • New DSX OS software components cover switch configuration, cluster-readiness testing, rack management and bare-metal provisioning.

  • Meta is adopting NVIDIA Spectrum-6 Ethernet switches for larger Vera Rubin AI infrastructure.

  • NVIDIA and Google announced Falcon Transport support on future ConnectX-10 SuperNICs and BlueField DPUs. That support is a forward-looking announcement.

The context

Why it matters

TideReach assessment: Standardized interfaces could let suppliers reuse compatible power, cooling and rack designs instead of solving each integration independently. That matters when electricity delivery and heat removal constrain expansion alongside accelerator supply. Operators will still need to validate complete systems: open interfaces do not guarantee that every component interoperates or remove reliance on NVIDIA hardware. Manufacturing and efficiency benefits remain vendor claims until demonstrated in actual deployments.