AI hardware · Layer

Networking and interconnect

Training a frontier model uses thousands to tens of thousands of accelerators that must act as one machine. The links between them, inside a server and across the data center, increasingly determine real-world throughput.

Key facts

Inside a node, proprietary links like NVIDIA's NVLink connect GPUs at very high bandwidth.
Across racks, InfiniBand and high-speed Ethernet carry traffic between nodes; the choice affects scaling efficiency.
Optical transceivers and switches are a fast-growing and supply-constrained part of the build.
At frontier scale, communication overhead between chips, not the chips themselves, often caps how fast a model trains.

Where it bottlenecks

Switch silicon and optical transceivers are supply-tight, and network topology can bound cluster scaling more than accelerator count.

Who dominates it

NVIDIABroadcomMarvellArista

Companies in this layer

Broadcom

United States

Supplier of networking switch silicon and a partner for custom AI accelerators (ASICs).

Marvell Technology

USA

Designs and supplies custom AI networking silicon, switch ASICs, and PAM4 DSPs for data center interconnects

All companies across the chain

Explore other layers

GPUs and AI accelerators High-bandwidth memory (HBM)Chip foundries and fabrication Advanced packaging Lithography and fab equipment AI servers and systems Data centers Cooling and thermal management Supply constraints and bottlenecks

Back to the AI hardware map

Let's get to work

Want to build with AI for real?

Beyond the explainer, we design, secure, build and run production AI. Tell us what you have in mind.

Start a project See the offer