ASML Just Priced the EUV Ceiling — and the AI Compute Trade Has to Pick Volume vs. ASP

On every line that matters for Q3, ASML printed at or above guidance and consensus. Revenue of €11,662M sat inside the company's €11-12B range and above the €11.72B Visible Alpha consensus; net income of €4,960M cleared the €4,051M consensus by more than 22%; basic EPS of €12.95 was the highest in the company's history, surpassing the prior €24.73 annual record on a per-quarter basis. The Installed Base Management line — field upgrades, service, refurbishments — printed €3,289M, roughly €400M above guidance and now running at a >50% growth pace for2026.

Two things changed in the Q3 update that did not show up in the headline beat. First, ASML confirmed it is investigating capacity beyond the previously communicated 30% increase for 2028 — language it did not need when the original 30%-on-top-of-30% curve was assumed sufficient. Second, CEO Christophe Fouquet disclosed that some customers are already asking for orders in 2029 and even 2030, well past the planning horizon the company is willing to publicly commit to. The combination — short-dated supply overshoot plus long-dated demand pull — is the textbook signature of a hardware-constrained market, not an inventory cycle.

Three mechanisms inside ASML are pulling more wafers out of each tool, which softens the volume ceiling without lifting it. The NXE:3800F is moving from 230 wafers per hour to 260 wafers per hour, with first systems shipping in small quantities in 2027 and the bulk of the impact in 2028+. Immersion throughput is being lifted to 330 wafers per hour in 2027. YieldStar optical metrology is qualified for 14Å logic and 1d DRAM with a 60% productivity improvement. Each lever adds effective capacity without selling another system, which is why Installed Base Management — the upgrade-and-throughput line — grew faster in Q3 than net-new system sales did.

A capacity-bounded monopoly does not have to compete on price; it has to compete on rationing. With ASML unable to deliver every EUV system every fab wants in 2027, three things become true simultaneously: the company can extract price for early delivery slots, customers pay higher tool revenue per wafer to secure capacity, and the value shifts toward the installed-base upgrades that lift wafer output per existing tool. The first is a one-time ASP win; the second is the structural answer; the third is the recurring annuity ASML has been building all year.

The volume side belongs to TSMC, which controls the largest share of 2027 EUV demand and reportedly pushed back on ASML's 10% tool price hike in mid-2026 before ultimately settling into a 2026 capex raise to $60-64B and an expected 2027 capex run-rate near $80B per Bank of America estimates. The ASP side belongs to the memory complex, where SK Hynix's $8B EUV commitment, Micron's record Q4 FY2026 gross margin of 86.8% on HBM-led revenue of $54.2B, and Samsung Electronics's planned High-NA volume ramp in 2028 are all paying up for tool priority. ASML's pricing power is being tested in real time on both sides of the same fab.

Tokyo Electron closed Oct 14 down 7.15% — the worst session of the regional equipment complex — even though its own Q1 FY2027 (April-June 2026) net sales of ¥732.3B beat consensus and printed +33.3% year-over-year. The market is treating the AI capex curve as if2026 was the peak. Lasertec, the sole-source supplier of actinic EUV mask blank and patterned-mask inspection, closed up 2.84% on the day but trades at a 47x trailing P/E and is being priced for the High-NA ramp rather than today's low-NA volume. Both stocks are reacting to the supply ceiling ASML just confirmed.

Short term, the binding catalyst is TSMC's Oct 15 Q3 print — the first chance the market gets to verify whether 2027 capex actually lands in the $76-85B Bank-of-America range or settles below it. TSMC's Q2 2026 capex of NT$396B (~US$12.7B) annualized already implied a $50B+ pace; a raise in the print confirms ASML's \"demand exceeds supply\" framing, and a hold or cut breaks it. Tokyo Electron's October 30 Q2 FY2027 release is the second test, because Tokyo Electron's quarterly revenue mix exposes whether SK Hynix and Micron are pulling forward EUV-adjacent orders or letting them slip into 2028.

Long term, the trade-off the print has now exposed becomes the dominant variable. If ASML succeeds in lifting 2028 capacity beyond the30% step-up it is investigating, the supply ceiling relaxes and the bull case reverts to volume on a flat ASP — which is what Tokyo Electron is currently being priced for. If ASML cannot physically add capacity at the rate customers want, the ASP lever compounds and the Installed Base Management line becomes the more important number to watch than system revenue — which is the bull case for the equipment sub-tier. The Q3 print did not resolve which world investors are in; it confirmed which axis they have to choose on.