Skip to content
AI supply chain term

Spine-leaf network

A 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.

What it means

Spine-leaf, a form of Clos topology, organizes a data center into leaf switches that servers plug into and spine switches that interconnect the leaves, with each leaf linked to every spine. This gives uniform, low latency and many parallel paths between any two servers — exactly what large training and inference clusters need when traffic can flow between any pair of machines. It sits at the physical-structure layer of the scale-out fabric, above the switch silicon and below the software that binds machines into one cluster. It is a lever because a facility can grow simply by adding leaves and spines without redesigning the network; it is a constraint because switch bandwidth, port count (radix) and cabling all set hard limits on how far a single flat fabric can scale before extra tiers are needed.

Why it matters to investors

Spine-leaf is the concrete shape of most AI cluster networks, so switch and system vendors that supply these tiers benefit as fabrics expand. Arista Networks (ANET) and Cisco (CSCO) supply the switches, while contract builders like Celestica (CLS) and Accton Technology (2345.TW) manufacture the boxes that fill the racks.

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

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.Network fabricThe high-speed network that ties accelerators together into a cluster.Scale-out vs scale-upConnecting many machines into one cluster over a network fabric.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.

See Spine-leaf network 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.