Skip to content
AI supply chain term

RDMA

RDMA (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.

What it means

RDMA moves data straight between the memory of two machines without the processors copying it through the operating-system networking stack, which sharply lowers latency and CPU load for the huge, frequent transfers that AI clusters make. It is the technology that turns a room full of servers into one usable scale-out compute resource, and it runs over InfiniBand or over Ethernet — the latter as RoCE, RDMA over Converged Ethernet. Sitting in the fabric layer, RDMA is what makes distributed training and large inference deployments efficient rather than bottlenecked on the network. It is a lever because it unlocks the full bandwidth of expensive interconnects; it is a constraint because it demands lossless, carefully tuned networks, which shapes the choice between InfiniBand and specialized AI Ethernet.

Why it matters to investors

RDMA is the reason a cluster's expensive interconnect actually delivers usable bandwidth, and the InfiniBand-versus-Ethernet choice around it steers spend across vendors. Arista Networks (ANET) and Hewlett Packard Enterprise (HPE) build RDMA-capable Ethernet networking, while cloud operator Oracle (ORCL) and system builder Accton Technology (2345.TW) deploy it at scale.

Companies on this part of the chain

Arista NetworksANETOracle CorporationORCLHewlett Packard EnterpriseHPEAccton 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

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

See RDMA 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.