Semicap tool-price inflation: capex dollars ≠ capacity

One-line summary: TSMC's CEO names tool-price inflation as one of only two reasons for a ~$10B capex raise — which means a rising foundry capex number is partly price, not capacity, and the inflation is accruing to the equipment vendors as pricing power.

The insight

Foundry capex is normally read as a capacity proxy: bigger capex → more wafers → more supply → the bottleneck eases. Asked what drove TSMC's 2026 capex from $52–56B (January) to $60–64B, cc-wei gave exactly two reasons, and only the first is capacity:

"the most important reason is because of demand continue to increase, and we feel the pressure from the customer... The second reason is inflation. Now we buy the tools with inflation price."

That is a first-party statement that part of the capex increase buys no incremental wafers at all. Two consequences follow, and they point in opposite directions for different names:

1. For the buyer (TSM), capex intensity rises without proportional capacity relief. The supply response to the AI shortage is weaker per dollar than the headline capex implies — which extends the bottleneck rather than resolving it. This is a quiet reinforcement of tsmc-capacity-shortfall-and-pricing-power and hbm-cowos-as-binding-bottleneck from an unexpected direction: the fix is more expensive than it looks.
2. For the vendors (ASML, AMAT, LRCX, KLAC), it is pricing power, disclosed by their largest customer. The customer with the most negotiating leverage on earth is publicly conceding it is paying inflated tool prices. That is the picks-and-shovels-leading-edge-fab-buildout thesis stated from the paying side of the invoice.

The pricing power has a visible boundary. Wei on High-NA EUV: TSMC works with ASML "to make it more suitable for manufacturing in terms of the cost and in terms of the maturity" and "always consider that technology maturity and the cost, and whether we use it or not." Tool vendors can inflate price; they cannot force adoption of a tool whose cost/maturity doesn't clear TSMC's bar.

Evidence

• cc-wei in 2026-07-16-earnings-tsm-q2-fy2026 — the load-bearing quote, in answer to "what is driving this [capex raise]?": "Well, simply put, the most important reason is because of demand continue to increase, and we feel the pressure from the customer to drive TSMC, not drive actually, to cooperate with TSMC for the capacity increase. That's one of the major reason. The second reason is inflation. Now we buy the tools with inflation price."
• The magnitude of the raise it explains — wendell-huang in 2026-07-16-earnings-tsm-q2-fy2026: full-year 2026 capital budget raised "to be between $60 billion and $64 billion." Per the analyst framing on the same call, January guidance was $52–56B and April was "closer to $56B" — an ~$8–10B raise, of which an unquantified share is price.
• No bottleneck on the supply of tools themselves — wendell-huang in 2026-07-16-earnings-tsm-q2-fy2026: "We always collaborate closely with the tool suppliers well in advance to prepare the capacity, whether it is a strong upcycle or downcycle... We do not foresee any bottlenecks to our capacity expansion plans." So this is price inflation on an available supply, not scarcity rationing.
• The adoption boundary on vendor pricing power — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026 on High-NA: "TSMC make it clear that we work with ASML and try to make it more suitable for manufacturing in terms of the cost and in terms of the maturity. We always consider that technology maturity and the cost, and whether we use it or not."
• The analyst asked directly and got a non-answer — Jeff Su put the question to Wei as "Is it because tool vendors are increasing the price?" and Wei's full reply was "It's all AI related. Everything." The split between price and volume in the capex raise is therefore not disclosed.

Implications

• ASML, AMAT, LRCX, KLAC — pricing power confirmed by the largest customer. Reinforces picks-and-shovels-leading-edge-fab-buildout.
• TSM — capex intensity rising faster than capacity; a margin and ROIC headwind that compounds with the disclosed 2nm (3–4pt) and overseas-fab (2–3pt → 3–4pt) gross-margin dilution.
• Second-order: every fab-buildout thesis that treats a capex number as a capacity forecast is now over-stating the supply response by an unknown factor. This applies to us-fab-capacity-bottleneck and the Arizona expansion arithmetic.

Contradictions / tensions

• The split is unquantified, and that is the whole weakness. Wei names inflation as reason two of two but gives no share. Without it, "capex dollars ≠ capacity" is directionally sourced and numerically empty. Do not size a position on this.
• Wei has an incentive to blame tool inflation for a capex raise that would otherwise read as chasing demand at any price — an interested framing, though an odd one to invent.
• Counter-pressure on vendor pricing: TSMC's demonstrated willingness to defer High-NA on cost grounds shows the pricing power is bounded by adoption discretion.

Open questions

• What share of the $8–10B raise is price vs. volume? (No source quantifies it.)
• Do ASML/AMAT/LRCX gross margins in H2 2026 corroborate price capture, or is TSMC's "inflation" mostly construction, labor, and materials rather than tools?

Valuation snapshot

Last refreshed 2026-07-20 (pre-open; marks are the Friday 2026-07-17 close, markets closed over the weekend). Every price fill tagged twelvedata. Mkt cap / Fwd P/E are not in the Twelve Data free tier — TSM's are sourced from stockanalysis.com on 2026-07-17 (tagged stockanalysis); the vendors' were not re-sourced (no figure shown rather than a fabricated one).

The irony of the day is worth recording: on 2026-07-16 the largest customer in the industry publicly conceded it is paying inflated tool prices, and every tool vendor fell.

Forward-looking outcomes (12-month)

Bull case (for the vendors) — H2 2026 vendor gross margins corroborate price capture: the load-bearing quote is as good as evidence gets in this project — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026, asked what drove the capex raise: "the most important reason is because of demand continue to increase... The second reason is inflation. Now we buy the tools with inflation price." That is the customer with the most negotiating leverage on earth publicly conceding it pays inflated prices — the picks-and-shovels-leading-edge-fab-buildout thesis stated from the paying side of the invoice. And it is price on available supply, not scarcity rationing: wendell-huang, same source, "We do not foresee any bottlenecks to our capacity expansion plans." If ASML/AMAT/LRCX gross margins expand in H2 2026, the claim converts from a quote to a number. Implied price: AMAT +25–35%; ASML +15–25%. Cited: 2026-07-16-earnings-tsm-q2-fy2026.

Base case — the direction is right, the split is never disclosed, and it changes nobody's model: the tool-inflation leg stays a qualitative reinforcement of a picks-and-shovels thesis that already stands on volume. Vendors capture some price inside a WFE cycle heading toward $160–185B by 2027; TSM wears a modest additional ROIC drag on top of the dilution it has already quantified. Implied price: vendors track the WFE cycle (see picks-and-shovels-leading-edge-fab-buildout base case); TSM +10–20%. Cited: 2026-07-16-earnings-tsm-q2-fy2026.

Bear case — "inflation" turns out to mean construction, labour and materials rather than tools: Wei named inflation but never said tool inflation was the bulk of it, and when cc-wei was asked point-blank by Jeff Su — "Is it because tool vendors are increasing the price?" — his entire reply was "It's all AI related. Everything." A non-answer. If H2 vendor margins are flat, the most likely explanation is that TSMC's inflation is fab construction and labour, and this concept dissolves. Implied price: no vendor downside from this leg failing — it is additive-only; TSM unaffected. Cited: semicap-tool-price-inflation.md (contradictions section).

Currently undervalued vs base case? No — and this page says so about itself, which is the point of keeping it. The concept's own Contradictions section is unusually blunt and it governs the answer: "The split is unquantified, and that is the whole weakness. Wei names inflation as reason two of two but gives no share. Without it, 'capex dollars ≠ capacity' is directionally sourced and numerically empty. Do not size a position on this."

That verdict is unchanged by the day's move. The vendors are 11–29% off their highs and the disclosure landed in their favour, but this page cannot tell you how much of an ~$8–10B capex raise is price versus volume, and without that the tool-inflation leg adds no sizing information to the picks-and-shovels-leading-edge-fab-buildout thesis it sits under. Route any actual position through that page, where the WFE volume case is quantified. What this concept legitimately does is bound two other pages:

• It makes tsmc-capacity-shortfall-and-pricing-power more durable, not less — the supply response to the AI shortage is weaker per dollar than the headline capex implies, which extends the bottleneck. The fix is more expensive than it looks.
• It marks down every capex-as-capacity forecast in the book — including the $100B Arizona arithmetic in us-fab-capacity-bottleneck — by an unknown factor.

Catalyst path:

• ASML Q2 (July), AMAT Q3 FY2026 (August), LRCX/KLAC (August) — do gross margins corroborate price capture? This is the one question that resolves the page, and all four report inside 60 days.
• TSMC Q3 2026 (October) — whether the $60–64B capex holds, and whether anyone gets Wei to split price from volume.
• Any High-NA adoption decision at TSMC for 1.4nm/A14 — the test of whether vendor pricing power is bounded by adoption discretion, as Wei implies.

Related

• picks-and-shovels-leading-edge-fab-buildout
• tsmc-capacity-shortfall-and-pricing-power
• us-fab-capacity-bottleneck
• cowos-packaging-capacity-crunch
• tsmc
• cc-wei
• wendell-huang