AI isn’t one trade. It’s a supply chain — chip designers, the fabs that build them, the memory feeding them, the buildings housing them, and the power plants keeping them on. Each sector below moves on its own.
It breaks AI into 18 parts of one supply chain and shows how each moved — today, this week, this month and this year. Open a sector to see the companies in it and what that part of the chain actually does. The point is to see which parts are moving differently from each other, rather than treating “AI” as one thing.
The same 18 sectors, ranked over three windows. A sector can lead the week and trail the year — which is the point of showing all three.
Equal-weighted average of the companies tracked in each sector, measured to the same close as the rest of this page. Reporting, not recommendations.
Databases, monitoring and orchestration — the layer AI applications are actually built on.
The chemicals, filters and subsystems a chip fab physically cannot run without.
The fabs that physically manufacture every AI chip — leading-edge and specialty nodes.
Stacking memory onto processors (CoWoS, HBM) — the physical bottleneck on AI chip supply.
The software and licensed blueprints used to design every AI chip before a fab ever touches it.
The processors that actually run AI models — the reason every other sector on this map exists.
The firms that bolt chips, memory and cooling into finished racks and ship them.
Unglamorous physical plumbing — and every AI server rack is full of it.
The machines that print chips. A handful of firms, no substitutes, multi-year lead times.
Lasers and transceivers moving data between racks at light speed.
The handful of buyers whose capex budgets fund everything else on this map.
HBM, DRAM, flash and hard drives — where the data an AI model trains on physically sits.
Transformers and high-voltage gear — the multi-year wait between a power plant and a data center.
Getting electricity into a rack and heat out of it — the constraint on how dense AI can pack.
The switches and links that let thousands of GPUs behave like one machine.
The landlords and rental fleets of the AI era — buildings full of GPUs, rented by the hour.
Nuclear, gas and alternative power feeding the electricity demand that AI data centers created.
The contractors pouring concrete and pulling conduit for the buildings themselves.
Tickers shown are the biggest one-month movers in each sector, not the full holdings list. Data: daily close · Reporting, not recommendations.