Google, Meta, Microsoft and Nvidia Unveil Grid Resilience Specification for AI Data Centers

At the Open Compute Project (OCP) Global Summit on October 12, Google, Meta, Microsoft and Nvidia jointly released a grid resilience specification requiring large-scale AI facilities to stay connected during grid disturbances and recover to near full load within less than two seconds.

The specification was among several AI infrastructure open standards announced at the summit. Nvidia also opened up its Vera Rubin rack and compute tray designs, contributing its MGX architecture to OCP so system makers can build infrastructure for next-generation AI factories. The updated MGX 1.2 specification adds rack-level blind-mate liquid cooling, interoperable rack interconnects, liquid-cooled 5000A busbars, and native 19-inch device support, enabling deployments of up to 250kW per rack and simplifying assembly, testing, maintenance and deployment.

Nvidia also released version 0.7 of the Open Data Center Specification with AMD, Google and Meta, which unifies power, cooling and physical layout designs for AI factories. In addition, Nvidia introduced four open-source software components for DSX AI Factory, covering switch configuration, GPU cluster stress testing, rack firmware management and bare-metal deployment.

Meta also announced plans to adopt Nvidia's Spectrum-6 102.4Tb/s Ethernet switches to build Vera Rubin AI factories on open networks. Nvidia and Cisco unveiled the USFP pluggable optical module standard for 400Gb/s network architectures, and Nvidia and Google said future ConnectX-10 SuperNICs and BlueField DPUs will support Falcon Transport to improve reliability and bandwidth utilization of Ethernet AI networks.