Skip to content
AI supply chain term

InfiniBand

InfiniBand is a high-speed, low-latency networking standard used to connect servers and GPUs into a single AI training cluster.

What it means

InfiniBand is a networking technology built for high-performance computing that connects many servers with very high bandwidth and very low, predictable latency. It offloads data movement from the CPU using RDMA, letting one machine read another's memory directly, which keeps thousands of GPUs working in lockstep during training. In the AI supply chain it is a leading scale-out fabric — the wiring that binds a whole facility into one usable compute resource. Its lossless design and tight latency make it a favorite for the largest training clusters, but it is a more specialized, often single-vendor ecosystem than standard Ethernet. That makes fabric choice a strategic decision: InfiniBand's performance versus Ethernet's openness and scale is one of the defining trade-offs in AI data-center design.

Why it matters to investors

The fabric is a genuine chokepoint because it determines how large and efficient a cluster can be, and it has become a battleground between InfiniBand and emerging AI-optimized Ethernet. Networking, systems, and optical suppliers are exposed to which standard the hyperscalers ultimately standardize on.

Companies on this part of the chain

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

See InfiniBand 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.

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.