Skip to content

The AI supply chain, layer by layer

Explore the layers of the AI buildout, from the chips that train models to the power and cooling that keep data centers running.

Explore

The AI supply chain, layer by layer

Chips

Accelerators and processors designed for AI workloads, made in leading-edge semiconductor fabs.

Read more

How it works

AI runs on a stack of physical infrastructure. Each layer depends on the ones below it, so spending on AI models flows down into chips, memory, networking, buildings, power and cooling.

How to use it. Select a layer to see what it includes and how it connects to the rest of the stack.

Why it matters. Looking at the whole chain shows where spending goes beyond the most familiar chip companies, which is the idea behind bill of materials investing.

FAQ

Frequently asked questions

What are the layers of the AI supply chain?

Chips and memory, networking, servers, data center buildings, power and electrical equipment, and cooling, with software and applications on top.

Why does AI need so much power?

AI chips draw far more electricity than standard servers, so data centers need new generation, grid connections and electrical equipment.

What is the bottleneck in the AI buildout?

It shifts over time. Recent constraints have included advanced chips, high bandwidth memory, power availability and grid equipment such as transformers.

What is bill of materials investing?

An approach that maps every component a buildout needs and looks at the companies supplying each piece, rather than only the most familiar names.

Learn more

Embed this tool

You can add this tool to your own website free of charge. Copy the code below; the tool stays up to date and links back to this page.

For educational purposes only. This material is not investment, legal or tax advice, and it is not a recommendation or offer to buy or sell any security. Figures from third-party sources are believed to be reliable as of the dates shown but are not guaranteed. Investments involve risk, including possible loss of principal.