About this site

Trillions of dollars are flowing into AI data centers, and most of that money ends up buying physical things: chips and the machines that package them, cold plates and chillers, transformers with multi-year queues, racks, cables, concrete and copper. Yet when you ask a simple question — who actually makes this part, and is there a public way to own it? — the answer is scattered across filings, trade press, partner lists and conference slides.

This site collects those answers in one place. It is a parts-first catalog: the unit of the site is the physical component, not the ticker. 130 part categories across 11 layers — from the GPU die out to the grid interconnection — with 1,202 companies attached to them through 1,862 placements. 682 of those companies have listed equity across 33 exchanges, each shown with its venue, ticker and a nightly price snapshot.

Who it's for

First, investors — retail and professional — asking "who makes X, and what's the public play?" Second, engineers, procurement teams, analysts and students who want to understand how the machine fits together. The site tries to serve the first group without ever talking down to the second.

What makes it different

• Two views of one graph. Part pages list who makes the part; company pages list what the company makes. Same data, two directions.
• Honest sourcing. Every placement carries a confidence tag — disclosed, reported, or inferred — rendered in exactly those words. Tags never upgrade without a source, and unknown stays blank.
• The physical story. The rack teardowns, the supercomputer breakdowns, the four chains, the scales matrix, the world map and real facilities connect the catalog to actual buildings and racks.
• Living data. Nightly jobs refresh prices, classify industry news against the catalog, pull Taiwan monthly revenue filings, and extract customer-concentration disclosures from SEC filings.
• Open access. The whole catalog is downloadable as SQLite, every page has Markdown and JSON mirrors, and llms.txt documents the stable IDs.

What it is not

Not investment advice, not real-time market data, and not affiliated with or endorsed by any company it covers. Placements tagged reported or inferred are leads, not confirmed supplier relationships — the sources & methodpage explains the discipline behind that in detail.

Corrections

The most valuable contribution is a correction with a public source — especially from people inside these companies. If we have you on the wrong part, or missing from the right one, email chiptogrid@simbasystems.com — details on the contact page.

The fine print

Terms of use · Privacy policy. The short version: informational reference only, no accounts, and Google Analytics for traffic measurement under Google's consent mode: it sets cookies only if you choose Accept all in the consent prompt.