At the OCP Global Summit, Nvidia also released the MGX 1.2 specification to open Vera Rubin rack and compute tray designs that support power levels up to 250 kW. The company announced new collaborations, including a common transport standard for Google's Falcon Transport on ConnectX 10 SuperNICs and new DSX open source software for cluster management. These updates complement the recent USFP pluggable optics standard created with Cisco, which supports 400 Gb/s per lane and delivers a 1.3x increase in power efficiency over 200 Gb/s per lane modules.
Tuesday, October 20, 2026
Story
Meta Deploys Nvidia Switches to Scale AI Clusters Beyond 100,000 GPUs
AI8 days ago
Nvidia is expanding its AI infrastructure footprint as Meta integrates the company's Spectrum 6 Ethernet switches into the Vera Rubin data center architecture. These switches provide 102.4 Tb/s of switching capacity and allow Meta to link more than 100,000 GPUs while reducing the number of required network tiers. This adoption secures Nvidia's place in Meta's hardware stack even as Meta continues to develop its own MTIA chips, opening further revenue opportunities for the chipmaker beyond GPU sales.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.