Skip to content
AI supply chain term

AI Ethernet (RoCE)

AI Ethernet is standard Ethernet tuned for AI clusters, using RoCE to carry RDMA traffic with the low latency and near-zero loss that training needs.

What it means

AI Ethernet refers to Ethernet networks adapted for AI workloads, typically via RoCE (RDMA over Converged Ethernet), which lets machines read each other's memory directly over ordinary Ethernet links. Efforts like Ultra Ethernet aim to give Ethernet the lossless, low-latency behavior that AI training demands while keeping Ethernet's open, multi-vendor ecosystem and huge supply base. In the AI supply chain it is the concrete wiring of most scale-out fabrics — the spine-leaf network that lets every machine reach every other quickly. It is the main open challenger to InfiniBand: a lever for hyperscalers who want to avoid single-vendor lock-in and reuse familiar Ethernet gear at data-center scale. Switch silicon, optics, and network-card design all determine how well Ethernet can close the remaining latency gap.

Why it matters to investors

Because Ethernet has many suppliers and enormous scale, a shift of AI fabrics toward RoCE and Ultra Ethernet would broaden who captures networking spend beyond a single incumbent. Merchant switch-silicon and systems vendors are the most direct beneficiaries to watch.

Companies on this part of the chain

Arista NetworksANETCiscoCSCOCelestica Inc.CLSAccton Technology2345.TW

Named to show where the term sits in the AI supply chain — research, not advice, and never a recommendation to buy or sell.

Related terms

RDMARDMA (remote direct memory access) lets one computer read from or write to another computer's memory directly over the network, bypassing the CPU and operating system to cut latency and overhead.InfiniBandInfiniBand is a high-speed, low-latency networking standard used to connect servers and GPUs into a single AI training cluster.Spine-leaf networkA spine-leaf network is a two-tier data-center wiring pattern in which every access ('leaf') switch connects to every core ('spine') switch, so any server can reach any other in a small, predictable number of hops.Switch siliconSwitch silicon is the specialized ASIC inside a network switch that decides where each packet goes and forwards it between servers at line rate — the engine of the data-center fabric.

See AI Ethernet in the live AI chain.

THE ENTITY maps every constraint onto one live model — which part is tight now, who owns it, and who gets squeezed when it moves. Plain-English reads you can check.

Explore the features
All glossary terms

THE ENTITY is an educational read on the AI supply chain — research, not investment advice. It explains how the chain works and who sits where, never price targets or buy/sell calls.