Where a Fab Dollar Goes
Every dollar of foundry capital spending is somebody else's revenue. When TSMC raised its 2026 capex guide to $60–64 billion — eight billion above the prior midpoint, announced the same week ASML committed to 30% more EUV capacity — it published a forward order book for the entire equipment chain. Roughly three quarters of a fab dollar flows to wafer fab equipment, the rest to construction, facilities and materials, and each layer of that flow has listed names attached. Reading the flow map tells you who gets paid; watching whether the tape pays them tells you what the market believes about the cycle.
What it is
Foundry capex splits into two streams. The larger — historically 70–80% — buys process equipment: the machines that print, deposit, etch, measure and test. The remainder builds the shells around them: cleanrooms, utilities, gas and chemical delivery, and the specialty materials that flow through a running fab. A capex raise is therefore not one announcement but a cascade of purchase orders with a known distribution, arriving at the equipment makers over the following quarters as orders, then shipments, then installed revenue. This lag is the point: a foundry's spend line guides the chain's income statements two to four quarters forward, which is why the market treats the capex guide — not the EPS beat — as the load-bearing number of a TSMC print.
Who gets paid
The first tier is lithography — the single largest line item, effectively one name for leading-edge EUV. Behind it, the broad wafer-fab-equipment basket: deposition and etch, then metrology and inspection, which scale overproportionally on new-node ramps because tighter geometries need more measurement per wafer. Then test, and — increasingly its own category — advanced packaging, where the CoWoS bottleneck gets debottlenecked with dedicated bonding, dicing and inspection toolsets.
Why it matters
The flow map turns one company's guidance into a testable forecast for a dozen others. When the spend line rises 15% at the midpoint, the chain's forward revenue rises with a lag — mechanically, contractually, with little room for interpretation. And the confirmation runs both ways: a toolmaker adding capacity (ASML's 30% EUV addition for 2027) needs a customer committing wafers; a foundry raising spend needs tools to buy. When both sides move in the same week — as they did in July 2026 — the build-out has confirmed itself from both ends of the supply chain. What the map cannot tell you is what the market will pay for it: equipment stocks discount the cycle, not the quarter, and the gap between confirmed flow and refused payment is itself information.
How to read it
Three reads. First, sequence: litho and deposition orders lead a node ramp, metrology scales through it, test and packaging arrive late — a capex raise touches the layers in order, not at once. Second, concentration: the more a raise is about advanced nodes and packaging, the narrower the beneficiary set; a shell-and-facilities raise (new fabs, new regions — the $100 billion Arizona commitment) spreads further into construction and materials. Third, and most important: watch whether the tape pays the flow. If the chain rallies on a foundry raise, the market believes the cycle; if it sells one — as it did on July 16, 2026 — the market is either fighting an ownership argument that overrides fundamentals, or it is telling you it believes the spend more than the return on it. The capex cliff is the standing risk on the far side of every raise.