ChipsThe story, in brief
OpenAI built a networking protocol with AMD, Broadcom, Intel, Microsoft, and NVIDIA to fix AI supercomputer bottlenecks
OpenAI just cut AI supercomputer networking layers in half. Here's why that matters for everyone else's capex.
The KeyNews take
Why it matters
OpenAI and five hardware giants (AMD, Broadcom, Intel, Microsoft, NVIDIA) built MRC, an open-source networking protocol that slashes GPU interconnect complexity and power consumption. It's already live on Stargate—signaling a fundamental shift in how AI infrastructure scales beyond 100K GPUs.
The key facts
5 to knowMRC protocol reduces switch layers from 3-4 to 2 for 100K+ GPU clusters
Cuts power and infrastructure costs for large-scale AI supercomputers
Already deployed on OpenAI's Stargate supercomputer
Open-source collaboration: AMD, Broadcom, Intel, Microsoft, NVIDIA
Solves AI supercomputer networking bottlenecks at scale
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: OpenAI has teamed up with AMD, Broadcom, Intel, Microsoft, and NVIDIA to develop MRC, an open source network protocol that sends data across hundreds of paths simultaneously between GPUs. Instead of three or four switch layers, MRC needs only two to connect over 100,000 GPUs, cutting both power and…