ASML Holding N.V. (NASDAQ: ASML) — The Only Toll Booth on the AI Highway, Priced for 2030’s Tolls

Ticker: ASML (NASDAQ) — the U.S.-listed line of the Euronext Amsterdam ordinary shares; the OTC grey-market line ASMLF is the same share, economically identical 1:1 (USD-quoted). The analysis applies to the ordinary shares regardless of trading line. Company: ASML Holding N.V. — Veldhoven, the Netherlands Sector / Industry: Technology — Semiconductor Capital Equipment (Wafer-Fab Equipment / Lithography) Report date: 2026-07-18 — UPDATE of the 2026-06-09 report · Currency: Reports in EUR under U.S. GAAP; quoted price in USD (EUR/USD ≈ 1.15, stated wherever used) Primary sources: Q2 2026 Form 6-K (filed 2026-07-15, accession 000162828026048235 — EX-99.1 press release, EX-99.2 investor presentation, EX-99.3 U.S.-GAAP statements, EX-99.4 statutory interim report); Q2 2026 earnings call (2026-07-15); FY2025 Form 20-F (filed 2026-02-25); SEC EDGAR CIK 0000937966

⚡ Kimi’s Take

Kimi’s own subjective opinion and general information; not investment advice. The body of this report (Sections 1–15) carries no position, no price target, and no recommendation.

Verdict: HOLD the business, AVOID the entry — UNCHANGED from the 2026-06-09 call, and the stock has spent five weeks demonstrating why: earnings caught up to the price, but the price never got cheaper. Accumulate only on a ~20–30% drawdown (roughly $1,200–1,400, ≈25–30× raised FY2026 EPS). Tag: “The only toll booth on the AI highway — the tolls just went up, and so did the price of the booth.”

Changed vs. the prior report: NO — same stance, re-anchored numbers. On 2026-06-09 the call was “HOLD the business, AVOID the entry — accumulate only on a 25–35% drawdown.” Since then ASML printed the strongest quarter in its history and raised FY2026 guidance a second time (€36–40B → €43–45B sales, +16% at midpoint; gross margin 51–53% → 54–56%), committed to +30% Low-NA EUV capacity for 2027 with the order book “close to fully covered,” disclosed explicit pricing power, and landed High-NA’s first high-volume logic product at Intel. The stock went nowhere (−1.7%; the print itself round-tripped in three sessions). That non-reaction is the thesis in miniature: the forward P/E de-rated from ~52× to ~38× entirely through estimate revisions, not price — yet the stock still sits at the 95.4th percentile of its own valuation history, and the price no longer capitalizes the 2030 model; it capitalizes an upward revision of it (Section 10). The 2026 guide midpoint (€44B) already touches the 2030 model’s low end four years early, so “delivering the plan” no longer clears the bar — the market is pre-paying the plan’s upgrade at the June 2027 Capital Markets Day. Two consecutive massive beat-and-raises (April, July) were sold into. Nothing in that combination argues for relaxing entry discipline; the discipline is simply re-anchored to higher earnings.

The framing is unchanged: a quality-compounder-at-a-price inside a crowded semis regime — not a falling knife, not a value opportunity, and not a momentum vehicle to ride blind. The factor read supports this precisely: the +133.5% trailing year is factor-carried, not alpha-carried (all-factor model R² = 0.83; loadings Netherlands +1.81, Semiconductors +1.19, Market +1.06; idiosyncratic vol 19.2%/yr), with the Momentum loading absent and Quality +0.63 — a real business carried by a real sector wave, not junk momentum. But the wave is at a statistical extreme: the semis industry factor sits at z = +2.42 (252d), tech at z = +3.02 (63d), short-term momentum is cooling (price below its 21-EMA; momentum factor 3d z = −1.51), and both of this year’s good prints faded within sessions. ASML is a 1.8-beta expression of a crowded trade that has stopped rewarding good news — a sizing/timing fact, not a business-quality fact. What flips me bullish: the price, not the print — a drawdown into ~25–30× raised FY2026 earnings (≈$1,200–1,400), or confirmation that the “under investigation” 2028 capacity leg is both committed and absorbed, which would make today’s price defensible on delivery rather than on hoped-for model revisions. What flips me bearish: a slip against the Q3 €11–12B guide — the first hard checkpoint of the largest shipment ramp in company history — or a formal export-control escalation (a new rule or investigation out of the June Lutnick EUV-to-China allegation) that strands the guided ~€8.8B China year. Conviction: medium-high.

Changes since 2026-06-09

This is an update report. The prior memo’s central claim — a durable monopoly priced for perfection — was tested by the strongest fundamental window in the company’s history and by a flat stock. The short version: the business outran even the prior bull case; the valuation did not get cheaper; the disclosure got worse.

What moved the thesis (all FACT unless noted):

• FY2026 guidance raised a second time, massively. €36–40B / 51–53% GM → €43–45B / 54–56% GM (Q2 2026 6-K, 2026-07-15); midpoint +15.8%. Q3 guided at €11.0–12.0B (+46–60% YoY) at 55–57% GM. The prior report’s bear-falsification test — “2026 lands at/above the top of €36–40B” — is now management’s base case: the bottom of the new range is €3B above the top of the old one.
• Q2 2026 beat on every line. Sales €9,326.5M (+21.3% YoY, above the guided high end), GM 54.0%, operating income €3,456.1M (37.1% margin), net income €2,917.6M, EPS €7.59. IBM (service/upgrades) €2,761.7M, +31.8% YoY, ~€300M above its own guide — the beat’s engine.
• Capacity commitment escalated a full year further out. 2026 Low-NA EUV output ~65 units; 2027 +30% to ~85, “close to fully covered with orders”; 2028 a further +30% (~110) “under investigation” with “significant” orders already in hand — order visibility Dassen said the company “hasn’t enjoyed in many years.” Immersion DUV follows the same +30%/+30% path (~130 → ~170 → ~220).
• Pricing power became explicit. Dassen: “the current environment provides more flexibility for pricing than what you would have had in different days… we’re executing on that” — a genuine tonal shift from Q1’s “our pricing is not based on the squeeze” (Reuters: “ASML has room to raise prices,” 2026-07-15).
• High-NA crossed from lab metrics to production proof. Intel Foundry is running ASML EXE High-NA on select 18A layers for Core Ultra Series 3 in high-volume manufacturing — the first HVM logic product on High-NA (ASML/Intel press releases, 2026-07-15). 4–5 EXE revenue recognitions guided for 2026 (3 done in H1).
• Bookings disclosure was discontinued. No bookings or backlog figure in either 2026 release; last prints: Q4 2025 bookings €13.2B, backlog €38.8B. The prior report’s #1 monitorable now exists only as management adjectives.
• The export-control guidance-bandwidth sentence was dropped in the same quarter the Lutnick EUV-to-China allegation surfaced (2026-06-18/19, denied by ASML) and the Netherlands joined the U.S.-led Pax Silica initiative (2026-06-23). Zero export-control mentions on the Q2 call; not one analyst asked.
• The stock de-rated on earnings, not price. $1,777.77 (2026-06-09) → $1,747.58 (2026-07-17), −1.7%, −12.6% off the 2026-06-30 all-time high ($1,999.96). Forward P/E ~52× → ~38×; own-history composite 96.4th → 95.4th percentile. Good news is being absorbed, not rewarded.

What was confirmed: the monopoly (no competitive event in the window; 100% EUV share); the IBM annuity thesis (+31.8% YoY, guided >30% — now the margin engine); the margin trajectory toward the 2030 model (54.0% actual, ~56% guided for H2); customer capex transmission (TSMC raised FY2026 capex to $60–64B the day after ASML’s print; Micron >$25B; SK Hynix ~$20.5B; ~50% HBM expansions); capital-return discipline (interim dividend +17.5% to €1.88; ~€1.0–1.1B/quarter buyback run-rate under the €12B program).

What was falsified: the prior bear case’s 2026 leg (below). Nothing else of consequence — the China-normalization path (36.1% → 29.1% → ~20% guided) is running ahead of schedule, and the valuation warning stands with one notch less extremity.

Prior report’s falsification tests — status:

📈 Stock Price Action — Five-Year Event Map

Five-year arc (FACT): from the October 2022 trough ($351 intraday / $366.50 close) the stock roughly 5.7×'d to an all-time intraday high of $1,999.96 on 2026-06-30, then backed off to $1,747.58 (2026-07-17 close), −12.6% off the high. The 52-week range is $679.89–$1,999.96; the stock is +133.5% over twelve months and +63.9% YTD 2026. The dominant repricings, oldest to newest (moves are FACT; attributed drivers are INTERPRETATION with the underlying events FACT):

Cycle narrative: (1) The 2022 drawdown was a macro de-rating meeting a genuine cyclical downturn, capped by the first U.S. export-control package — the two forces (multiple compression + policy) that still define the downside. (2) The January 2024 bookings print marked the moment the market began pricing AI as an ASML order-book story. (3)–(4) The 2024 crashes show the stock’s asymmetric sensitivity to order intake and export headlines — a −16.3% single-day move on a bookings miss is the live template for the bear case. (5)–(6) 2025 whipsawed on macro (tariff-pause rally) and management’s own caution (the July refusal to confirm 2026 growth), bottoming at $680 — the launch point of the entire melt-up. (7) The September 2025–June 2026 repricing was sector-wide and factor-carried: TSMC’s capex guide, record backlog, and two guidance raises re-rated the whole WFE complex (ASML, AMAT, LRCX all printed record highs on 2026-06-17). (8) The July 2026 fade is the second consecutive “good print, weak tape” — incremental good news absorbed rather than rewarded. Unresolved drivers (OPEN QUESTION): the +9.5% day on 2026-06-11 and the −7.8% day on 2026-06-23 have no confirmed company-specific catalyst; the latter coincided with a hawkish-Fed tape and the Pax Silica headline — treat both as unattributed macro/positioning.

1. Executive Summary

ASML Holding N.V. is the world’s monopoly supplier of extreme-ultraviolet (EUV) lithography systems and the dominant supplier of deep-ultraviolet (DUV) immersion systems. Lithography — printing circuit patterns onto silicon — is the single most critical and most expensive step in semiconductor manufacturing, and ASML is the only company on Earth that can build the machines required to make the most advanced chips. No leading-edge logic node (3nm, 2nm and below) or advanced DRAM can be manufactured at volume without ASML’s EUV tools. This is as close to a true, sole-source monopoly on an enabling technology as exists in any major industry.

The financial signature of that position strengthened further this quarter. In FY2025 ASML generated €32.7 billion of revenue (+15.6%), a 52.8% gross margin, €11.3 billion of operating income, €9.6 billion of net income, and ~€11.0 billion of free cash flow on only €1.6 billion of capex. The first half of 2026 extended every one of those trends: H1 2026 revenue €18.1 billion (+17.2% YoY), gross margin 53.5%, operating margin 36.6% (+160bps), net income €5.7 billion (+22.2%), EPS €14.74 (+23.9%). Q2 alone printed €9,326.5M of sales (+21.3%), a 54.0% gross margin, and a 37.1% operating margin — all multi-year bests. Nearly 30% of Q2 revenue (IBM, €2.76B, +31.8% YoY) is now high-margin, annuity-like service and upgrades — the recurring stream that cushions new-system cyclicality and that drove both the revenue and the margin beat.

The demand backdrop has moved from strong to extraordinary. ASML raised FY2026 guidance for the second time this year, to €43–45 billion of sales at a 54–56% gross margin — a €44B midpoint that already touches the low end of the company’s 2030 model (€44–60B) four years early. Management quantified the year for the first time: EUV systems +45% (~65 shipments), memory systems +75%, logic foundry systems +25%, IBM +30%, China ~20% of sales. For 2027, Low-NA EUV capacity rises ~30% to ~85 units and is “close to fully covered with orders”; a further +30% for 2028 (~110 units) is under investigation with orders in hand. High-NA crossed from engineering milestone to production proof: Intel is running EXE tools on 18A layers for Core Ultra Series 3 in high-volume manufacturing. And for the first time, management spoke openly about raising prices.

The debate, therefore, remains what it was in June — price and cyclicality, not quality — but both sides of it have hardened. The stock trades at $1,747.58 (2026-07-17), −12.6% off the all-time high set on 2026-06-30, a ~$671 billion market cap, the 95.4th percentile of its own ten-year valuation history, 54.9× trailing earnings and ~38× the raised 2026 estimate. Embedded in that price is no longer merely the 2030 model but an upward revision of it: at a 9% required return, today’s price needs 2030 EPS of €56–64 at a 35–40× exit multiple, and base-case delivery leaves the stock ~35% underwater in present-value terms (Section 10). Meanwhile the risks have rotated rather than retired: China stepped down to ~20% of guided revenue under an export regime that grew louder (the Lutnick allegation, Pax Silica) even as ASML quietly dropped the export-control caveat from its guidance; the FY2026 raise is heavily H2-weighted, requiring the largest shipment ramp in company history; and the quarterly bookings series — the cleanest independent check on demand — has been discontinued. This memo takes no position on the security; it lays out the monopoly, the accelerating economics, and the expectations the price now embeds.

2. Business Overview

ASML designs, manufactures, markets, and services photolithography systems for the semiconductor industry, plus the associated metrology/inspection, computational-lithography software, and lifecycle services. Lithography is the patterning step in chip fabrication: light is projected through a mask (reticle) onto a photoresist-coated silicon wafer to define the transistor and interconnect features of a chip. The smaller the feature ASML’s tools can resolve, the more advanced the chip a customer can build. ASML does not make chips; it makes — and is, for EUV, the only company that makes — the machines that make the chips.

Why this matters so disproportionately: chip advancement is fundamentally a story of printing ever-smaller features, and feature size is governed by the wavelength of light used to print it. The industry spent two decades stretching 193nm DUV light to its physical limits before the only way forward became EUV at 13.5nm — requiring light generated by vaporizing tin droplets into plasma, reflected off mirrors polished to near-atomic flatness, in a vacuum, at nanometer precision. ASML is the only company that solved this, after two-plus decades and well over €10 billion of cumulative investment, in partnership with Zeiss (optics) and its own Cymer subsidiary (light source). The result: the entire global advance of computing — every leading-edge AI chip, every flagship smartphone processor, every advanced DRAM die — passes through a tool that exactly one company on the planet can build. That is the business in one sentence.

Product lines (FACT, per FY2025 20-F and Q2 2026 6-K):

• EUV lithography (TWINSCAN NXE / EXE platforms). 13.5nm extreme-ultraviolet light for the smallest features on leading-edge logic and advanced DRAM. ASML is the world’s only manufacturer of EUV systems — a literal monopoly. “Low-NA” EUV (NXE, 0.33 NA) is in high-volume manufacturing across all leading-edge fabs; “High-NA” EUV (EXE, 0.55 NA) is the next generation, and crossed a defining threshold on 2026-07-15: Intel Foundry qualified EXE on select 18A layers and is producing Core Ultra Series 3 processors — the first high-volume logic product built with High-NA. Low-NA tools sell for roughly €180–220M; High-NA for ~€350–400M. Q2 2026 EUV system sales were €3.8B (including one High-NA unit).
• DUV lithography (TWINSCAN immersion and dry). 193nm systems for mature and mid-critical layers. ASML is dominant (~80–90% of immersion) but competes here with Nikon and Canon. Q2 2026 non-EUV system sales were €2.8B. DUV is the principal product sold into China.
• Metrology & inspection. YieldStar optical metrology and HMI e-beam systems, increasingly bundled with the lithography roadmap (the “holistic lithography” strategy).
• Installed Base Management (IBM) — service & field options. Maintenance, spare parts, and performance/productivity upgrades across the multi-thousand-unit installed base — the recurring, high-margin annuity, and as of 2026 the fastest-growing line in the company.

Revenue model and segmentation (FACT, EUR):

System sales are large, lumpy, and capital-cycle-sensitive (recognized on shipment/acceptance of individual machines). IBM is recurring and grows structurally with the installed base — it rose 26% in 2025 and 28.1% in H1 2026, is guided to grow >30% for FY2026, and was the engine of the Q2 beat (~€300M above its ~€2.5B guide, “driven primarily by additional upgrade business”). Q2 2026 system sales split 51% Logic / 49% Memory (Q1: 49/51). H1 2026 system sales by technology (FACT, statutory interim): EXE 3 units / €1,189M; NXE 29 / €6,709M; ArF immersion 40 / €3,332M; ArF dry 13 / €367M; KrF 65 / €773M; i-line 20 / €113M; metrology & inspection 104 / €362M. EUV (EXE+NXE) was €7.9B, +38.5% YoY, while DUV+metrology fell 9.6% — H1 was EUV-led, with immersion deliberately H2-weighted (management prepped Zeiss for a lower 2025 number and says it will “make up for that quite substantially in the second half”).

Customers and end-markets. ASML’s customers are the world’s chipmakers: foundries (TSMC), logic IDMs (Intel, Samsung), and memory makers (Samsung, SK Hynix, Micron). End-demand is downstream — AI accelerators, smartphones, PCs, data-center CPUs, automotive, industrial. ASML sells “picks and shovels” to the entire semiconductor gold rush rather than betting on any single chip architecture. The regional mix shifted markedly in H1 2026 (FACT, statutory interim): South Korea 39.2% of sales (memory super-cycle), Taiwan 28.3%, China 15.9% (from 24.1% in H1 2025 and 29.1% in FY2025), United States 10.7%.

How the machine makes money — the unit economics. A low-NA EUV scanner (NXE) carries an ASP of ~€180–220M, weighs ~180 tonnes, and takes months to install and qualify; a High-NA EXE roughly doubles that to ~€350–400M. ASML recognized 48 EUV units in FY2025 (44 in 2024) and shipped 91 lithography systems in Q2 2026 alone — a small number of very large transactions drives system revenue, which is why quarters are lumpy. The offset is DUV volume (hundreds of units at lower ASPs) and the IBM annuity. Critically, revenue is recognized largely on shipment/acceptance while customers pay substantial down-payments in advance (€17.4B of contract liabilities at end-H1) — the customer funds ASML’s working capital, a structural sign of pricing power few capital-goods businesses enjoy. New in 2026: management says the environment “provides more flexibility for pricing than what you would have had in different days” and that it is “executing on that” (Q2 2026 call; Reuters, 2026-07-15). With long order lead times, any pricing effect lands in 2027–28 ASPs, not 2026.

Holistic lithography — the bundling strategy. ASML sells not a scanner but a patterning system: scanner plus YieldStar metrology, HMI e-beam inspection, and computational-lithography software. The bundle raises value per litho layer and deepens lock-in — the customer’s yield is co-engineered with ASML’s tools — and it is the platform on which the fast-growing upgrade business is sold.

Logic vs. Memory balance. Q2 2026 system sales split 51% Logic / 49% Memory, but the H1 growth tells the real story of this cycle: memory system sales rose 53.2% YoY while logic system sales fell 10.0%, with memory systems guided +75% for the full year (DDR/HBM price strength, “multiple mega fabs”) and advanced-logic foundry systems +25% (implying a very large H2 logic acceleration). ASML is not a one-end-market story — but the 2026 cycle is, unusually, memory-led, and memory is the most procyclical of its customers.

Verdict: A focused, sole-source supplier of the single most critical tool in semiconductor manufacturing, with a growing recurring-service annuity layered on lumpy but extraordinarily profitable system sales. The model monetizes the entire industry’s race to smaller nodes — on customer-funded working capital, with a bundled software/metrology layer compounding the lock-in. New in 2026: the annuity is now the growth engine (IBM +28% H1, guided >30%), and pricing power has moved from implicit to stated.

3. Industry Dynamics

Structure. Semiconductor capital equipment (wafer-fab equipment, “WFE”) is an oligopoly of giants, each dominant in its own process step: Applied Materials and Lam Research (deposition/etch), KLA (process control), Tokyo Electron (coat/develop), and ASML (lithography). Within lithography the structure is the most concentrated of all: ASML is a monopolist in EUV and dominant in DUV immersion, with Nikon and Canon relegated to older DUV and niches. Lithography is also the most valuable step — the EUV tool is the single most expensive machine in a fab — so ASML captures a disproportionate share of the WFE profit pool. Nothing in the current window changed this structure; the one challenger narrative (the xLight free-electron-laser startup) remains pre-commercial.

Market size and growth. WFE is a >$130B market in 2026, and estimates are being revised up: SEMI forecasts record ~$139B equipment sales in 2026 and ~$156B in 2027; Citi’s bull case runs $145B (2026) → $200B (2027) → $250B (2028); Morgan Stanley raised 2026 WFE to ~$128B and 2027 to ~$150B; KLA expects >$140B in 2026 with 2027 growth exceeding 2026 (sources dated 2026-04 to 2026-06; see Section 15). Three secular drivers remain genuinely independent:

1. AI infrastructure (the demand accelerant). AI accelerators are built on leading-edge logic and require enormous quantities of HBM — the two most lithography-intensive products in the industry. The transmission evidence this window was unambiguous: TSMC raised FY2026 capex to $60–64B (from $52–56B) the day after ASML’s print, raised its revenue outlook to >40% growth, and called demand robust “through 2030” (2026-07-16/17); Samsung and SK Hynix are each targeting ~50% HBM capacity expansion in 2026 (SK Hynix capex ~$20.5B, +17%); Micron’s FY26 capex is >$25B with HBM4 sold out through 2026. ASML’s own order commentary — “extremely strong” H1 intake, backlog increasing, 2027 Low-NA “close to fully covered,” 2028 orders “two years in advance… we haven’t enjoyed in many years” — is management hypothesis, but it is corroborated by these customer-level facts.
2. Node migration / rising litho intensity (the structural escalator). Each new logic node and DRAM generation adds EUV layers; High-NA then adds another tier. Fouquet (Q2 2026 call): customers are investing at 3nm and 5/4nm, ramping 2nm, and “already planning investment to support the development of the 1.4nm nodes”; DRAM’s 1b/1c nodes use more EUV layers, and HBM “will require more wafer.” Litho spend per wafer rises with every node independent of unit growth. A new twist: TSMC is staying on low-NA extension (A14) rather than jumping to High-NA, which raises near-term NXE layer intensity even as Intel proves High-NA on 18A.
3. The installed-base annuity (the de-cyclicalizer). Every system shipped becomes a decade-plus stream of service and upgrade revenue. IBM grew 26% to €8.2B in 2025, +28.1% in H1 2026, and is guided >30% for FY2026 (~€10.7B implied); in Q2 2026 it was 29.6% of sales.

Management’s framing (Q2 2026): “continued strong momentum in customer demand… customers have revised their capital expenditure plans upward… entering into long-term agreements with their customers.” Note the tone evolution versus Q1: the slogans (“demand will continue to outpace supply,” customers “sold out”) were not repeated — they were replaced by harder substantiation (2027 order cover, 2028 orders in hand, capacity +30%/+30%). Substance strengthened while rhetoric cooled. ASML’s 2030 model (2024 Investor Day) targets €44–60 billion of revenue at a 56–60% gross margin; the FY2026 midpoint (€44B) now touches that range’s low end four years early, which is why the next Capital Markets Day — dated June 10, 2027 — matters: it is the venue where the model is likely to be revised up (INTERPRETATION).

Capacity as the industry’s binding constraint — and ASML’s answer. The signal of the quarter is that ASML itself is the bottleneck and is committing to relieve it: Low-NA EUV output ~65 units in 2026, +30% capacity for 2027 (~85), a further +30% under investigation for 2028 (~110); DUV immersion ~130 in 2026, +30% to ~170 in 2027, a further +30% (~220) investigated — all within the existing footprint (freeing cabins, relocating R&D tools, cutting cycle time; no new cleanrooms). A new campus breaks ground in 2026 but serves demand “beyond 2028.” Dassen added a productivity kicker: the 2027 mix shift from NXE:3600D to 3800E/F means the effective wafer-capacity uplift is ~45%, not 30%. The 2027–28 plans even incorporate expected demand from Elon Musk’s planned “Terafab” facility (Reuters, 2026-07-15) — a measure of how far out customers are now planning.

Barriers to entry (Greenwald). EUV combines all three genuine advantage types:

• Proprietary technology / supply advantage: ~three decades and well over €10B of cumulative R&D, plus a supply chain (Zeiss optics, Cymer light source, tens of thousands of components) no entrant can assemble. ASML owns Cymer and holds a 24.9% equity stake in Carl Zeiss SMT, the sole supplier of EUV optics.
• Economies of scale: ASML’s €4.7B annual R&D budget (~14% of sales) exceeds the entire equipment R&D of any would-be competitor. Scale is self-reinforcing: more tools → more service revenue and field data → faster roadmap → more tools.
• Customer captivity: once a fab is built around ASML tools, switching is impossible — the process is co-developed with ASML, and the installed base locks in a decade of service/upgrade dependence.

Regulation. The defining external factor remains export control. EUV has been effectively banned for export to China since 2019; advanced DUV immersion restrictions tightened from 2024. Since the prior report, the temperature rose without a new formal rule: on 2026-06-18/19 Bloomberg reported that U.S. Commerce Secretary Lutnick told senior ASML executives he is concerned an EUV machine may have reached China — ASML denied it (no EUV system or EUV-specific components delivered; the ~180-tonne tools are tracked, serviced, remotely monitored), and no public evidence of a breach has been presented (Reuters, 2026-06-18; TechCrunch, 2026-06-19). On 2026-06-23 the Netherlands joined the U.S.-led Pax Silica AI supply-chain initiative “despite disputes over China-bound exports by ASML” (Reuters, 2026-06-23) — reducing the odds the Dutch licensing regime diverges from Washington. A February 2026 bipartisan letter from eight U.S. lawmakers calling for a blanket country-level SME export ban remains live. Against that backdrop, ASML’s Q2 guidance dropped the sentence that had sized the guidance band to “accommodate potential outcomes of ongoing discussions around export controls” — and not one analyst asked about China on the call. Whether the removal signals reduced perceived risk or reduced disclosure caution is an OPEN QUESTION; this memo treats the policy risk as unchanged-to-higher even where the disclosure implies otherwise.

Capital-cycle read (Marathon). The industry is deep in a capital-cycle boom — customers are raising capex mid-year (TSMC +$8B at midpoint one day after ASML’s raise), memory makers are in synchronized ~50% HBM expansions with sold-out production, and sell-side WFE paths are being revised up across 2026–2028. Normally a boom invites supply, erodes returns, and mean-reverts. The critical distinction: ASML is the toll-collector, not a commodity capacity-adder. It faces no entrants; its monopoly suspends the competitive mean reversion capital inflows would otherwise cause. ASML’s capital-cycle risk is therefore not competitive erosion but its customers’ over-building — a downstream capex digestion showing up as a lumpy air-pocket in system orders. This window both pushes that risk further out (2027 order cover, 2028 orders in hand) and enlarges its eventual magnitude (the build is bigger, more synchronized, and memory-led — the most procyclical segment). Marathon’s rule applies: synchronized, price-justified capacity additions are what finance the next down-leg.

Profit-pool concentration. Within a fab’s equipment budget, lithography is the most expensive single category and, because ASML is sole-source in EUV, the one with the least price elasticity. As nodes advance, litho’s share of WFE spend rises, so ASML captures a growing slice of a growing pie — the structural reason its margins expand with scale while deposition/etch peers share their profit pools with strong rivals. The “agency” dynamic reinforces it: the chipmaker’s customers (Apple, Nvidia, the hyperscalers) effectively demand leading-edge nodes, and leading-edge nodes require ASML — the buyer cannot substitute away. Pricing power flows from indispensability, not negotiation — and for the first time, ASML is saying so out loud.

Verdict: Structurally excellent — among the best industry positions in global equities, and strengthened since June. Lithography is the most concentrated, highest-value, highest-barrier node of an oligopolistic, secularly-growing industry, and ASML owns its monopoly tier outright. What changed since the prior report is demand, and the direction is unambiguously up: ASML’s own guide +€7B at both ends, TSMC’s capex +$8B a day later, memory in a ~50% HBM expansion, WFE paths revised up across the Street, and a High-NA adoption proof-point landed. The only demand component moving down is China (36% → 29% → ~20%), more than replaced by AI-driven logic and memory elsewhere. The structural caveat is unchanged: ASML’s fortunes ride on its customers’ capex cycle, and this boom — like every boom — is building the inventory of its own eventual digestion.

4. Competitive Position

The moat, named. ASML’s competitive advantage is the strongest configuration in the Greenwald framework — economies of scale combined with customer captivity, reinforced by genuinely proprietary technology — applied to a monopoly product. In EUV there is no second source. This is not “differentiation” (which Greenwald shows does not protect profits); it is a structural barrier competitors cannot scale, because the technology, supply chain, patents, learning curve, and installed-base data all compound in ASML’s favor and cannot be replicated on any commercial timeframe. Nothing in the 2026 window touched this; the June–July news flow was about ASML extending the lead (High-NA’s first production use), not defending it.

Mechanism, sub-system by sub-system:

• Optics: EUV mirrors must be polished to atomic-scale flatness; Carl Zeiss SMT is the sole supplier, and ASML owns 24.9% of it. High-NA and low-NA optics are “totally different tools… not fungible” (Dassen) — each platform generation re-locks the dependency.
• Light source: EUV light is generated by vaporizing tin droplets with a high-power laser ~50,000 times per second. ASML acquired Cymer (2013) to own this. The roadmap (1,000-watt source demonstrated; 330 wafers/hour targeted at the start of the next decade) is itself a moat-deepening exercise.
• Systems integration: an EUV tool has ~100,000 parts and a supply chain of thousands of specialized vendors orchestrated only by ASML — and it is this integration capacity, not demand, that is the industry’s binding constraint (hence the +30%/+30% capacity program executed within the existing footprint).
• Installed-base lock-in: each tool ties the customer to a decade of service and upgrades; the “holistic lithography” bundle raises switching costs further. The Q2 2026 IBM line — €2.76B, +31.8% YoY, ~€300M above guide, “very high-margin components” driving the gross-margin beat — is this lock-in monetizing in real time.

The market-share and ROIC tests (Greenwald). EUV market share is 100% and has been since the product existed (Greenwald’s test: <2-point change over 5–8 years = formidable barriers; ASML has zero share loss). Sustained returns run far above the 15–25% threshold that signals real advantage — FY2025 ROE 59.9%, ROIC 38.8% (ROIC.ai), with H1 2026 operating margin of 36.6% (+160bps YoY) showing the economics still improving with scale. Both tests are passed decisively, from an identifiable and durable source.

Where the moat is narrower. In DUV, ASML is dominant but not sole — Nikon and Canon compete in dry and older immersion. DUV is a very good business but not a monopoly, and it is the segment most exposed to China and to mature-node commoditization. The H1 2026 mix illustrates the point: EUV system sales grew +38.5% YoY while DUV+metrology fell 9.6% — the monopoly tier is carrying the year.

Direct comparison vs. competitors. Nikon and Canon abandoned the EUV race a decade ago; neither can credibly re-enter. Their reasons are instructive: EUV required betting billions over twenty-plus years with no certainty of a working product, mastering tin-plasma sources and atomic-flatness optics, and orchestrating a supply chain that did not exist. Two profitable, competent optics companies concluded the investment was uneconomic — and were proven right that only one firm could amortize that spend across the whole industry’s demand. Within WFE, AMAT/Lam/KLA/TEL are excellent businesses in their own steps and do not compete with ASML in lithography. There is no competitor to ASML’s core product; the “competition” that matters is downstream — whether customers keep investing.

Pricing power — from structural to stated. The prior report inferred pricing power from structure (sole-source, pre-payments, queueing). The Q2 2026 call made it explicit. Dassen: “the current environment provides more flexibility for pricing than what you would have had in different days… we’re executing on that” (UBS Q&A); and “we believe the potential to capture a larger share of that value… gives you better pricing power. Those are the conversations that we’re currently having with customers” (Wolfe Q&A). This is a genuine tonal shift from Q1 2026 (“our model of pricing is not based on the squeeze”). With 2027 Low-NA “close to fully covered with orders,” ASML holds the strongest pricing hand in its history. Validation metric for coming quarters: EUV revenue per unit and gross margin versus mix — pricing effects lag by order lead times (INTERPRETATION; the commentary is FACT).

High-NA — the monopoly extender crosses a threshold. On 2026-07-15 ASML and Intel announced that Intel Foundry is using EXE High-NA EUV on select Intel 18A layers to produce Core Ultra Series 3 (“Panther Lake”) processors in high volume — the first HVM logic product on High-NA, at yields “matched to the NXE platform” (GlobeNewswire/ASML, 2026-07-15; Reuters, 2026-07-15). Fouquet called it “maybe the strongest sign so far that we’re getting there.” Four to five EXE systems are guided for 2026 revenue recognition (three done through H1). The competitive significance: High-NA adoption risk — the one technology-execution risk the prior report flagged — moved from “prove the lab metrics” (Q1’s SPIE data: 0.5M wafers, >80% availability) to “proven in production.” Note the adoption asymmetry: the first mover is Intel, not TSMC, whose A14 strategy extends low-NA — TSMC remains the swing buyer for High-NA unit volume in 2027–28 (OPEN QUESTION).

China’s domestic-litho effort (SMEE) as the only entry attempt. The one place an entrant is being willed into existence is China, where SMEE and state-backed programs are attempting domestic DUV (and, aspirationally, EUV) lithography under sanctions pressure — the Marathon “state-capitalism breaks the capital cycle” scenario. Realistically, SMEE is years behind in DUV and effectively nowhere on EUV (which needs the very Zeiss optics and Cymer sources that are export-controlled). The June Lutnick episode — a U.S. allegation that an EUV machine may have reached China, denied by ASML on logistics grounds — underscores the moat: even motivated state actors cannot field an EUV-class tool, and Washington’s anxiety is about leakage of ASML’s, not emergence of a rival’s. SMEE remains a long-dated risk to ASML’s China DUV revenue, not a threat to the EUV monopoly on any foreseeable horizon.

Verdict: A durable, identifiable, sole-source monopoly — one of the widest moats in the public markets, and wider than it was in June. The advantage shows up in 100% EUV share, ~60% ROE, a 54% gross margin rising toward 56%, customers who pre-pay and queue two years out, and — new this quarter — management openly exercising pricing power. The caveats are unchanged: the moat’s depth (EUV) is narrower in scope than the revenue’s breadth (DUV + service), and growth is hostage to customers’ capex decisions, not to any competitive threat.

5. Growth History and Forward Opportunities

Historical growth (FACT, EUR): Revenue grew from €21.2B (2022) to €27.6B (2023, +30.2%) to €28.3B (2024, +2.6%) to €32.7B (2025, +15.6%) to €18.1B in H1 2026 alone (+17.2% YoY). The 2024 near-plateau was a digestion year (post-2023 China DUV pull-forward and a memory downturn); 2025 reaccelerated on AI logic and recovering memory; 2026 is tracking to ~€44B at the guidance midpoint — +34.6% versus FY2025, one of the fastest growth years in company history. The longer arc is a low-teens revenue CAGR across the past decade as EUV moved from introduction to the backbone of leading-edge manufacturing.

Composition of recent growth (FACT). The 2025 increase (€4.4B) was logic-led plus a 26% jump in IBM. H1 2026 rotated: memory system sales +53.2% YoY (logic systems −10.0%), EUV +38.5%, IBM +28.1%. The FY2026 guidance decomposes (management statements, Q2 2026 call — treat as management arithmetic): EUV systems +>45% (~65 Low-NA shipments, implying ~€16.8B); non-EUV systems ~+25% (~130 immersion shipments, ~€16.1B); IBM +>30% (~€10.7B); memory systems +>75%; advanced-logic foundry systems +>25%; China ~20% of total sales. The internal cross-check (16.8 + 16.1 + 10.7 ≈ €43.6B) lands inside the €43–45B range.

Forward opportunities:

1. The AI capacity build-out (near term, upgraded). Q3 2026 guided at €11.0–12.0B (+46–60% YoY) at 55–57% GM; implied Q4 is €12.9–15.9B (vs €9.7B in Q4 2025). The raise is heavily H2-weighted — management attributes it to demand plus an immersion supply-chain re-acceleration — making H2 execution the binding variable (INTERPRETATION on FACT arithmetic).
2. 2027 visibility (the new leg). Low-NA EUV capacity +30% to ~85 units, “close to fully covered with orders”; immersion +30% to ~170; the NXE:3600D→3800E/F mix shift implies ~45% effective wafer-capacity uplift. For 2028, a further +30% EUV (~110) is under investigation with “significant” orders already received — visibility “we haven’t enjoyed in many years.”
3. EUV layer intensity (structural). Each new logic node (2nm ramping, 1.4nm in planning) and DRAM generation (1b/1c) adds EUV layers; HBM adds wafer volume on top.
4. High-NA EUV (the next monopoly, now in production). Intel 18A/Core Ultra Series 3 is the first HVM product; 4–5 EXE recognitions guided for 2026; ~2× low-NA ASP. Caveat: TSMC, the swing buyer, is extending low-NA on A14 first.
5. Installed Base Management (the compounding annuity). Guided >30% for 2026 after +28.1% in H1; ASML is “significantly expanding” the upgrade portfolio. The steadiest, highest-margin growth vector.
6. Pricing (the new lever). Explicit pricing flexibility with long order lead times — upside to 2027–28 ASPs and the gross-margin path that current guidance does not yet reflect (INTERPRETATION).
7. The 2030 model — and its likely revision. The 2024 Investor Day targets €44–60B revenue and 56–60% GM by 2030. With 2026 already at the range’s low end, the June 10, 2027 CMD is the venue where the model gets refreshed — plausibly upward (INTERPRETATION).

Decomposing the 2030 bridge. The €44–60B 2030 target now implies only an 8–11% revenue CAGR from the raised 2026 base (€44B) — a band that narrowed dramatically because the 2026 raise pulled the base forward roughly four years. The planks: (1) EUV units and ASP — ~65 (2026) → ~85 (2027) → ~110 (2028, investigated), layered with High-NA at ~2× ASP and, new, explicit pricing; (2) DUV resilience — immersion ~130 → ~170 (2027), re-accelerating in H2 2026 as the Zeiss supply chain catches up; (3) IBM — compounding at ~30% off the installed base. The gross-margin path to 56–60% rests on EUV/High-NA mix, better-priced H2 configurations, IBM strength, and fixed-cost absorption — H2 2026 is guided at ~56%, i.e., the 2030 margin model’s low end is also being reached four years early. None of these planks requires share gains — ASML already has 100% of EUV; they require the industry to keep investing, which is the cyclical bet.

EUV unit trajectory as the leading indicator. EUV units recognized: 44 (2024) → 48 (2025) → ~65 planned (2026) → ~85 capacity (2027) → ~110 under investigation (2028). Each unit is ~€180–220M (more for High-NA) plus a decade of attached service, so the unit count is the cleanest forward gauge of both system revenue and the future IBM annuity. The 2028 decision — confirm, defer, or cancel the investigated +30% — is the single most informative future data point on the demand cycle (flagged for CMD 2027-06-10 and interim call commentary).

The disclosure problem: no more bookings. ASML discontinued quarterly bookings and backlog disclosure with the 2026 reporting year (FACT — neither 2026 release carries the figures; last disclosed: Q4 2025 bookings €13.2B, backlog €38.8B). What remains: management adjectives (“extremely strong,” “close to fully covered”), capacity-commitment announcements, contract-liability movements (€17.4B at end-H1, down €1.9B from YE2025), and shipment actuals. The order book is now less independently verifiable than at any point in recent history — a genuine analytical loss (INTERPRETATION).

Quality of growth. Still high-quality growth: organic, funded by internal cash flow (the share count is falling), at expanding margins (operating margin +250bps YoY in Q2), underpinned by a monopoly rather than reversible share gains. The principal qualification remains cyclicality, plus a new one: the 2026 acceleration is memory-led (memory systems +75% guided), and memory is the segment with the deepest history of synchronized over-building. The growth is real; its composition is the most procyclical of this cycle.

Verdict: High-quality, monopoly-underpinned, organically-funded growth — upgraded in magnitude since June, cyclical in cadence, and newly dependent on an H2 execution ramp. The secular drivers (AI, node migration, High-NA, service annuity, pricing) are real and largely supply-constrained at the customer level; 2027 is closer to booked than any forward year in recent memory. The risk to the growth thesis is timing and digestion, not durability — and with the bookings series gone, the first sign of that risk will arrive as a shipment miss, not an order decline.

6. Financial Quality

ASML’s financial statements are among the cleanest and most powerful in large-cap technology. Every figure below is FACT from the FY2025 20-F or the Q2 2026 6-K (EUR, U.S. GAAP; the statutory interim is EU-IFRS and shows H1 net income €438.6M lower on development-expenditure capitalization — we use U.S. GAAP throughout for consistency with the prior report).

Profitability and margins.

Margins are not only high but expanding with scale — the most important test of a real moat (Greenwald: economics must improve with scale). Q2 2026 gross margin hit 54.0% — the best since at least 2023 and above guidance — “primarily due to the contribution of very high-margin components within our Installed Base Management business.” Operating margin rose faster than gross margin (37.1%, +250bps YoY) because R&D (+9.4%) and SG&A (+1.3%) grew far below sales (+21.3%): operating leverage is showing up exactly as a monopoly at scale should. H1 opex grew +5.5% against +17.2% sales. H2 gross margin is guided at ~56% (Dassen) — mix (more immersion, better-priced EUV “230 configurations”), IBM strength, and fixed-cost absorption.

Quarterly trend (FACT, U.S. GAAP):

Cash generation — the down-payment caveat is now load-bearing. Free cash flow (company definition: OCF − capex − intangibles) was €9.1B (2024) and €11.0B (2025) — ~95%+ conversion of net income on capex of ~€1.6B (~5% of sales): a capital-light monopoly. But H1 2026 free cash flow was negative €1.29B (operating cash flow −€482.5M against €5.67B of net income) — the mirror image of Q4 2025’s +€11.4B down-payment surge. The H1 drain is dominated by two working-capital moves: accounts receivable +€4.2B (fast shipment/recognition late in Q2; €7.25B of AR at quarter-end) and contract liabilities −€1.9B (down-payment burn as backlog converts to revenue). Quarterly OCF/FCF is noise for this company; annual FCF is the only meaningful lens. But the magnitude is larger than prior years, and the FY2026 cash story now requires a sharp H2 recovery — historically the pattern (Q4 collections), currently an assumption doing real work, since H1 dividends + buybacks (€3.73B) were funded entirely from the balance sheet.

Working capital and the prepayment signal. Inventories are large (€11.7B) — the cost of building €350M machines with year-plus cycle times — but the offset is contract liabilities (customer prepayments): €17.4B total at end-H1 (€13.9B current + €3.6B non-current). Customers pre-fund ASML’s working capital; a business whose customers pay in advance for a place in the queue is exhibiting textbook pricing power. The H1 decline in that balance (down-payments converting to revenue faster than new ones arrived) is the mechanical shadow of the shipment ramp — worth watching, not yet a warning.

Balance sheet (Jun 28, 2026). Total assets €50.2B; total equity €21.8B (43.5% of assets); cash + short-term investments €7.6B (down from €13.3B at Dec-2025); long-term debt €1.98B → net cash ~€5.6B (excluding the €1.7B current debt portion; ~€3.9B including it). The ~€5B H1 decline is the working-capital swing plus €3.7B of dividends/buybacks and €0.7B of debt repayment — not operational deterioration. Goodwill is €4.6B (legacy Cymer/HMI/Berliner Glas), modest relative to equity. Leverage is negligible, though the cash cushion is half what it was at year-end.

Returns on capital. FY2025: ROE 59.9%, ROIC 38.8%, ROA 19.4% (ROIC.ai, consistent with filing-derived figures) — franchise returns sustained over a decade, the Greenwald signature of genuine advantage. With FY2025 EBIT of €11.3B, ~17% tax (NOPAT ~€9.4B), and invested capital of roughly €16B (equity €21.8B less net cash €5.6B), after-tax ROIC sits comfortably above 30% — far beyond the 15–25% threshold, and on H1 2026 run-rates tracking at or above the FY2025 level. This business earns a monopoly return on capital, not a market return.

Quality-of-earnings checks. Net income tracks cash flow closely on an annual basis. Share-based compensation is trivial (€89.6M in H1, ~0.5% of sales); the share count is falling via buybacks (weighted-average basic 384.5M in Q2 2026 vs 388.2M a year earlier). Equity-method profit (€145.7M H1, mostly the Zeiss stake) is real economic income. The effective tax rate is a clean ~17.3% (FY2026 guided ~17%, newly quantified). One new cost item to monitor: Q2 opex came in above guidance on “estimated costs related to the technology and IT transformation,” primarily in R&D — under EU-IFRS, H1 R&D was +40.0% YoY (the U.S.-GAAP +5.7% excludes the capitalization difference). Size and duration are undisclosed (OPEN QUESTION for the Q3 call and the 20-F); it is the only new line muddying an otherwise pristine cost picture.

Multi-year trend read. Three patterns stand out across 2023→H1 2026. First, operating leverage is real and accelerating: cumulative revenue +31% since 2023 with operating margin up from 32.8% to 37.1% (Q2 2026), because opex compounds at roughly one-third the rate of sales. Second, mix is steadily enriching margins — IBM (the highest-margin line) grew 26% in 2025 and 28% in H1 2026 and funds the gross-margin step-up; within systems, EUV is displacing DUV. Third, the recurring layer is de-risking the model: IBM rose from €6.5B (2024) to €8.2B (2025) to a ~€10.7B implied 2026 — a larger and larger share of revenue is non-cyclical.

Peer cross-read. Against the WFE complex, ASML’s economics stand out specifically because of the monopoly. AMAT/Lam/KLA/TEL earn excellent but competed returns (gross margins high-40s to high-50s, operating margins high-20s to low-30s) shared with strong rivals. TSMC earns extraordinary margins but must sink $60–64B of 2026 capex (~40% of revenue) to defend its lead — a capital-intensive moat. ASML pairs comparable margins with capex of ~4.5% of sales and a literal monopoly, so its cash economics (FY2025 FCF/sales ~34%, ROIC ~39%) are arguably the best in the semiconductor value chain. Memory peers (Micron, SK Hynix, Samsung) sit at the cyclical, commoditized end — the volatility ASML’s toll-collector model is insulated from on the competitive axis (though not the demand axis).

Verdict: Exceptional financial quality — and improved since June. Economics improve with scale (54.0% GM, 37.1% operating margin, both multi-year bests), annual cash conversion is ~95%, ROIC exceeds 30%, the balance sheet is net cash, and earnings quality is high (trivial SBC, ~17% tax, no one-offs beyond the unquantified transformation costs). Two qualifications, both about timing rather than health: (1) H1 cash conversion was negative on working-capital swings — the H2 recovery is now a load-bearing assumption; (2) the €7B guidance raise is concentrated in H2 shipments and depends on supply-chain output re-acceleration and a China re-acceleration — the raise is management’s plan, not booked backlog cover (and with bookings disclosure gone, it cannot be independently checked). Nothing in the quarter weakens the financial-quality thesis; the disclosure loss removes the cleanest independent check on it.

7. Capital Allocation

Philosophy. ASML runs a disciplined, shareholder-friendly policy: fund the R&D roadmap that protects the monopoly, invest modestly in capacity, and return the bulk of free cash flow via a growing dividend and large buybacks. It carries net cash and uses no leverage to flatter returns. On the Marathon asset-growth test, ASML is firmly on the right side — it does not empire-build; it shrinks the share count (weighted-average basic shares −1.4% YoY in H1 2026).

R&D — the moat-maintenance budget. €4.7B in 2025 (14% of sales), €2.46B in H1 2026 (13.6%), self-funded — the most important capital-allocation decision the company makes, funding High-NA, source power, and the holistic roadmap. Note the new “technology and IT transformation” costs flowing through this line in 2026 — an efficiency reorganization (~1,700 coordination roles out, ~1,400 engineering roles in, announced January 2026) intended, per Dassen, to get “even more value out of the team” without headcount growth. Sound intent; unquantified cost.

Capital returns (FACT).

• Buybacks: H1 2026 repurchases of €2,077.5M (Q1 €1.0B + Q2 €1.08B, ~0.8M shares at ~€1,375 average) under the €12B 2026–2028 program (announced 2026-01-28; up to 2M shares for employee plans, remainder to be cancelled). ~€9.9B remains; the required pace is ~€1.0B/quarter and H1 ran at ~€1.04B — on track, neither accelerated nor slowed. Two tempering facts: the prior 2022–2025 program was also “up to €12B” but executed only €7.6B — the ceiling is not a commitment; and H1 2026 cash deployed (€2.08B) is roughly half of H1 2025 (€4.08B) — the heavy buying was done at the early-2025 trough (~$683); current execution is mechanical run-rate at record prices.
• Dividends: FY2025 total €7.50/share (+17%); final €2.70 paid in Q2 2026. The first 2026 interim is €1.88/share, +17.5% vs the €1.60 2025 interims, payable 2026-08-05 — the board is tracking another ~+17% dividend year (only the interim is declared; beyond that is extrapolation). Payout ratio remains modest (~26–30%), leaving ample headroom.
• The funding gap (the new fact of 2026): H1 2026 total cash returned was €3,733.0M (€1,655.5M dividends + €2,077.5M buybacks) against H1 free cash flow of −€1.29B — returns are currently funded from the balance sheet, which is why net cash fell from ~€10.6B to ~€5.6B in six months. A linearity artifact (Q4 2025 alone produced €11.4B of operating cash flow), not a funding problem — but the payout now leans on the H2 recovery, and the dry-powder cushion is half what it was if the cycle wobbles in 2027.

M&A. The acquisition history is small, strategic, and capability-driven — Cymer (2013), HMI (2016), Berliner Glas (2020) — plus the 24.9% Zeiss SMT stake and, in 2025, the €1.3B Mistral AI investment (~11% stake, an AI-software hedge rather than a capability deal). These are supply-chain-securing plays of the kind Greenwald endorses — the opposite of revenue-synergy deals that destroy value. No record of large, dilutive, ego-driven acquisitions.

Capacity investment — the right kind of “growth capex,” now larger. Capex (H1 2026 €808M, ~4.5% of sales annualized) funds capacity expansion — and the announced program (+30% Low-NA EUV and +30% immersion for 2027, a further +30% investigated for 2028, a new campus breaking ground in 2026 for beyond-2028 demand) is demand-pulled, pre-sold capacity into a queue, executed mostly within the existing footprint. Every unit of added EUV capacity is a unit of monopoly revenue plus a decade of attached service — the opposite of a commodity producer building into hoped-for demand. Capex will trend up into 2027–28 (INTERPRETATION); watch the cash-flow line for corroboration of the “no new cleanrooms” claim (modest capex = consistent).

Insider activity and incentive alignment.

• Insider transactions: N/A by structure (FACT). As a Dutch foreign private issuer ASML is exempt from Section 16 reporting — EDGAR shows zero Forms 3/4/5 across 2024–2026 — and Dutch AFM insider registers are not in EDGAR. The correct statement is “insider-transaction evidence is not available in the U.S. filing record,” not “no insider selling.”
• Remuneration (FACT, FY2025 20-F): STI is an annual cash bonus (target max 150% of base); the LTI 2025–2027 weights are relative TSR 25% + financial value drivers 35% + non-financial measures 40%, with TSR + financial ≥60% of any cycle and vesting capped at 200%. The non-financial leg includes a Technology Leadership Index — management is paid partly on maintaining the technology lead, an unusual and moat-relevant design (INTERPRETATION). Insider ownership is small (~0.8%), so alignment runs through compensation design rather than founder stakes.

The one critique — buyback timing — now better evidenced, in both directions. ASML is repurchasing stock at the 95th percentile of its own valuation history; buying at ~55× trailing earnings is value-neutral-to-destructive relative to buying in a drawdown. But the record is better than “procyclical”: the heaviest buying (Q1 2025, €2.59B) was done at the early-2025 trough, H1 2026 deployment halved versus H1 2025, and the prior program under-executed its ceiling (€7.6B of €12B) — ASML bought roughly twice as much stock when it was cheap and throttled to run-rate at the peak. That is price-insensitive rather than actively procyclical — mechanical, not valuation-driven — and mildly contra the prior report’s “running hot” framing. The dividend (+17.5% interim step-up, modest payout) remains the best-judged leg. The genuinely intelligent deployment this year is capacity: +30%/+30% into pre-sold, queue-backed demand at trivial capex.

Verdict: Excellent capital allocation — disciplined, shareholder-aligned, net-cash, no value-destructive M&A — with the timing quibble now better evidenced on both sides. Buybacks run at a mechanical ~€1B/quarter into a top-decile valuation, but the record shows larger buys at the 2025 trough and under-execution of the prior ceiling; the dividend is unambiguously well-judged; and the capacity program is the highest-return allocation on the menu. The watch-item is new: returns currently outrun free cash flow, funded from a net-cash cushion that halved in six months — safe under the H2 recovery assumption, thinner if the cycle turns.

8. Changes and Headwinds — Last Two Years

Strategic / operational developments:

• AI demand inflection → super-cycle (2024→2026). AI infrastructure investment turned the 2024 digestion year into a reacceleration that is still steepening. FY2026 guidance went €34–39B (January) → €36–40B (April) → €43–45B (July) with the GM band up 300bps to 54–56% — the €44B midpoint already reaches the low end of the 2030 model four years early. Management now quantifies the year segment-by-segment (EUV >45%, memory >75%, IBM >30%) and speaks of 2028 orders “two years in advance… we haven’t enjoyed in many years.”
• Capacity escalation. The output plan moved from “at least 60 Low-NA EUV in 2026, at least 80 in 2027” to ~65 → ~85 (+30%) → ~110 investigated (2028), with DUV immersion on the same +30%/+30% trajectory, all within the existing footprint; a new campus breaks ground in 2026 for beyond-2028 demand. ASML is confirming it remains the industry’s bottleneck and is committing capital to relieve it.
• High-NA commercialization — from lab to production. First HVM-ready EXE:5200B shipped April 2025; SPIE metrics in early 2026 (0.5M wafers, >80% availability); and on 2026-07-15 the first high-volume logic product — Intel 18A Core Ultra Series 3 — at NXE-matched yields. 4–5 EXE recognitions guided for 2026 (3 done). The prior report’s High-NA adoption risk retired one notch.
• Pricing power explicit. From “our pricing is not based on the squeeze” (Q1) to “more flexibility for pricing… we’re executing on that” (Q2). New in the company’s recent history.
• Bookings disclosure discontinued (2026). Last prints: Q4 2025 bookings €13.2B, backlog €38.8B. Order evidence is now qualitative commentary plus capacity announcements.
• Leadership. Christophe Fouquet (CEO since 2024) and Roger Dassen (CFO) — no changes in the window; strategy continuous. The January 2026 restructuring (~1,700 coordination roles out, ~1,400 engineering roles in) is flowing through as “technology and IT transformation” costs.
• Capital returns reset. €12B 2026–2028 buyback program; dividend +17% for 2025, interim +17.5% for 2026.

Headwinds and overhangs:

• Export controls / China — escalated in tone, unchanged in rule. EUV is banned to China; advanced DUV restricted since 2024. New in the window: the Lutnick EUV-to-China allegation (2026-06-18/19, denied by ASML, no public evidence), the Netherlands joining Pax Silica (2026-06-23), and a live February 2026 bipartisan letter seeking a blanket country-level SME ban. Meanwhile China revenue stepped down faster than modeled — 36.1% (2024) → 29.1% (2025) → 15.9% of H1 2026, guided to ~20% of FY2026 (~€8.8B) — and ASML dropped the export-control bandwidth sentence from its guidance. The prior report’s nuance holds: China demand is DUV-weighted, and the EUV profit core has zero China exposure by construction. The slower-burning concern is the China service base: curtailed new-tool sales eventually erode the attached high-margin annuity. And the guidance now embeds a re-acceleration — H1 at 15.9% implies H2 China of ~€5.9B (~2.4× the H1 run-rate) to land the 20% year — with no stated policy buffer (ASSUMPTION inside guidance).
• Customer concentration intensifying. The largest customer (TSMC) was 23.9% of 2025 sales; four customers = 61.2%. AI is concentrating demand further — TSMC’s capex raise to $60–64B is the same concentration, one node downstream. Korea alone was 39.2% of H1 2026 sales.
• Cyclicality / digestion risk — pushed out, and enlarged. 2024 demonstrated that a customer capex pause can flatten a year. The 2027 order cover and 2028 order book push the digestion risk further out; the synchronized memory build (three makers, ~50% HBM expansions, record prices) enlarges its potential magnitude. The 2028 capacity leg being merely “under investigation” is management’s own hedge against exactly this.
• Tariffs / trade environment. Section 232 review of the ~$143B semiconductor-equipment market was still pending as of April 2026; zero tariff mentions on either 2026 call, but unresolved.
• Execution risk (new). The FY2026 guide implies H2 sales of €24.9–26.9B, +45–56% H2/H2 — the largest shipment ramp in company history, dependent on the immersion supply-chain re-acceleration and the China re-acceleration. A slip lands the year at or below the low end.
• Cash-cushion drawdown (new, minor). Net cash halved to ~€5.6B in H1 on the working-capital swing and €3.7B of shareholder returns. Not a solvency issue; a thinner buffer for the next downturn.

Verdict: On balance the changes strengthen the thesis — more than they did in June. The AI inflection, capacity escalation, pricing power, and the High-NA production milestone are large positives that outweigh the (real, rising) China/export overhang. The headwinds remain concentrated in geopolitics, cyclicality, and now H2 execution — not competitive position. The honest counterweight: two of the positives (the raise, the capacity plan) are management plans rather than booked facts, and the disclosure environment for verifying them just got worse.

9. Risk Analysis (Risk Matrix)

The risks have rotated since June: execution and geopolitics are now first-order, while competitive and technology risks receded one notch. Every risk below is grounded in evidence from the Q2 2026 6-K, the Q2 call, the FY2025 20-F, or validated news reports (sources P1–P5, T1, N1–N23).

Risk synthesis. The dominant risks are cyclicality (medium likelihood, high impact) and geopolitics (medium likelihood, high impact), compounded by a starting valuation that leaves little room for either to disappoint. Execution risk is newly prominent because the FY2026 raise is H2-weighted. The good news: none of the top risks is competitive — the moat itself is not on the list. The bad news: a non-moat risk meeting a 95th-percentile multiple can produce a 30–50% drawdown anyway, as 2022 and 2024 demonstrated.

10. Valuation Discussion — Embedded Expectations

All figures below are reconciled to the Q2 2026 6-K and FY2025 20-F; FX convention is EUR/USD = 1.15. The price is $1,747.58 (2026-07-17 close) = €1,519.6; shares outstanding 384.1M; market cap €583.7B ($671.3B); net cash €5.6B; EV €578.1B (net cash is ~1% of market cap, so EV ≈ market cap). The standing rule applies: no price target, no recommendation — this is a decomposition of what the current price implies.

Current multiples (filing-based)

Key interpretation: ASML de-rated on earnings, not price. The forward P/E fell from ~52× to ~38× because FY2026 EPS estimates jumped from ~€29 to ~€40, while the stock price was flat. The own-history composite percentile fell only 1.0 point (96.4th → 95.4th) — the market paid up for the beat-and-raise rather than letting the multiple collapse. That is the central valuation fact: the price capitalizes an upward revision of the 2030 model, not merely delivery of the old one.

Comp set

*KLAC has split 10-for-1 since June; $212.75 × 10 = $2,127.50, flat vs June.

Read: Cross-sectionally ASML is no longer the most expensive equipment name — it sits between AMAT (~51× trailing) and LRCX/KLAC (~60×). Relative to its own history, ASML is now the least extreme of the equipment majors (95.4th vs 99.86–99.98th). In absolute terms it remains at the richest TTM multiple of its own decade — above the 2021 bubble-era peak.

Embedded expectations — what must be true at $1,747.58

Decomposition 1 — required 2030 EPS at a 9% required return. Assuming 4.45 years to end-2030 and ignoring the modest dividend, the required 2030 price is ~€2,230.

*Base case ~€48; 2030 model top ~€60–65 (ASSUMPTION).

Conclusion: today’s price only “works” at a 9% return if ASML delivers the top of the 2030 model and the market still pays 35–40× for it in 2030 — or if earnings exceed the entire current 2030 model. Because the 2026 guide midpoint (€44B) already touches the 2030 model’s low end, the price is implicitly underwriting a revision upward of the 2030 model at the 2027-06-10 Capital Markets Day.

Decomposition 2 — PV of scenario delivery at 9%.

Conclusion: today’s price ≈ the present value of the bull case. Base-case delivery at a still-premium 30× exit leaves the stock ~35% underwater in PV terms.

Decomposition 3 — reverse-DCF. Off a raised FY2026E FCF base of ~€13–15B, an EV of €578B requires a ~14–15% FCF CAGR for a decade, then 3% terminal growth. That is roughly double the WFE industry’s historical trend, from a cyclical peak, uninterrupted.

Scenario analysis through 2030

Anchors: FY2026 guide €43–45B / 54–56% GM; 2027 Low-NA EUV capacity ~85 units “close to fully covered”; 2028 +30% (~110 units) “under investigation”; 2030 model €44–60B / 56–60% GM. Share count assumed ~372M by 2030 on continued buybacks.

Bear mechanics: memory system orders gap in 2027 H2–2028 (precedent: 2023 industry-wide memory capex −19%); the 2028 capacity add is cancelled; China compresses to ~8–10% of revenue (~€3–4B) under a country-level SME ban; the multiple mean-reverts to a normal cyclical 20–28×. The bear case does not require the monopoly to break — only an ordinary semicap digestion meeting a 95th-percentile multiple with no asset-value floor. ASML has already done −52% (2022) and −64% (lifetime max drawdown), so the range is historically grounded.

What the market is pricing correctly: the monopoly, the demand cycle (TSMC raised capex to $60–64B the day after ASML’s raise; memory trio in synchronized ~50% HBM expansions), the margin trajectory, the IBM annuity, and 2027 order visibility.

What the market may be pricing incorrectly: the joint probability of all those good things happening simultaneously and uninterrupted for 4.5 years; the durability of the memory-led leg; the assumption that China settles at ~20% with no new rules; and the persistence of a 35–40× exit multiple into 2030.

Verdict: The stock is priced for an upward revision of the 2030 model, not merely its delivery. The forward multiple has de-rated on earnings catch-up, but the absolute valuation remains extreme. The asymmetry is clear: base-case delivery leaves meaningful downside in present-value terms; the bull case is required to justify the price.

11. Variant Perception

Consensus belief, as of the Q2 2026 print. ASML is a unique, must-own AI-infrastructure compounder. Demand outpaces supply, the 2026 guide was raised twice, 2027 is “close to fully covered,” High-NA is in production, and the stock’s modest pullback from the June high is a buying opportunity. The bull case is largely priced as base case.

Where consensus may be too low (upside variant).

1. 2027 estimates. Street 2027 numbers were set before the Q2 disclosures. Bottom-up capacity math (~85 Low-NA EUV units, immersion +30%, E/F mix shift adding ~45% effective wafer capacity, explicit pricing power) supports 2027 revenue of €50–55B vs the one visible post-print contributor model at ~€47–48B (2027 EPS ~€42.40). Consensus likely has to come up.
2. 2030 model revision at CMD 2027-06-10. With 2026 already at the 2030 model’s low end, an upward revision is highly probable — a dated catalyst the market cannot yet fully capitalize.
3. Pricing power. First explicit pricing commentary in years; with 2027 nearly booked, pricing effects land in 2027–28 ASPs — unmodeled upside to the GM path.

Where consensus may be too high (downside variant).

1. The H2 2026 ramp. Implied Q4 revenue of €12.9–15.9B is the largest shipment ramp in company history. It depends on immersion supply-chain re-acceleration and a China H2 re-acceleration to ~€5.9B (~2.4× H1). A slip lands the year at/below the low end — and at a 95th-percentile multiple, “in line with the old guide” is a miss.
2. Memory durability into 2028. The +75% memory-systems guide is the most procyclical leg at its most synchronized moment. The market models the capacity adds; it does not model the digestion (Marathon capital-cycle framing).
3. The multiple itself. The 95.4th-percentile composite and a semis factor at +2.42σ (252-day) mean the re-rating is statistically extreme. Persistence is not permanence.

Factor-positioning read. The trailing-year +133.5% return is factor-carried, not alpha-carried: all-factors model R² = 0.83; dominant loadings are Netherlands +1.81, Semiconductors +1.19, Market +1.06; idiosyncratic vol is only 19.2%/yr. Momentum loading is absent (ElasticNet zeroed it), Quality is +0.63 — this is a real business carried by a real sector wave, not junk momentum. But the wave is at a statistical extreme: semis factor z = +2.42 (252d), tech factor z = +3.02 (63d), short-term momentum cooling (3d z = −1.51). Two consecutive good prints (April, July) were sold into. The marginal price-setter is sector/factor flow, not stock-specific fundamental discrimination — ASML can de-rate on sector mean-reversion even if every company-specific thing goes right.

The three assumptions that matter most. (1) 2027–28 WFE does not digest. (2) China/export regime does not materially worsen. (3) The multiple stays 35–40× into 2030. The price requires all three; any one breaks and the PV math turns against today’s entry.

Verdict: The variant perception is two-sided and finely balanced. The bull variants (2027 estimates, 2030 model revision, pricing) are real and under-modeled. The bear variants (H2 ramp, memory digestion, multiple) are also real and under-priced. The tie-breaker is the starting valuation: at a 95.4th-percentile composite, the burden of proof rests on the bull.

12. Fact vs. Interpretation

13. Open Questions

1. What is the actual Q2-only China revenue split? The statutory interim discloses H1 geography only (China 15.9%). Bear-case mechanics use H1 vs FY-guide to infer H2; precision awaits the Q3 interim.
2. How large and how long are the “technology and IT transformation” costs? Q2 R&D ran above guidance on this new line; it is the only new cost item muddying an otherwise pristine picture.
3. What is the hard consensus for 2027/2028 EPS? The “consensus too low on 2027” claim rests on one contributor model plus bottom-up capacity math; a professional consensus pull would sharpen or falsify it.
4. Will the 2028 +30% Low-NA EUV capacity add be confirmed, deferred, or cancelled? This is the single most informative future data point on the demand cycle; flagged for the 2027-06-10 Capital Markets Day.
5. Does export-control risk escalate from rhetoric to rules? The Lutnick allegation, Pax Silica, and the bipartisan letter are live; ASML dropped the guidance buffer language while China is guided to re-accelerate in H2.
6. Can ASML actually ship the implied Q4 2026 revenue (€12.9–15.9B)? This would be the largest quarterly shipment in company history — a critical H2 execution checkpoint.
7. What will the 2030 model look like after revision at CMD 2027? With 2026 already at the old model’s low end, the revision is highly probable; the magnitude determines whether today’s price is early or full.
8. When will peer Q2 prints (AMAT, LRCX, KLAC, late July–August) refresh the comp table? Current peer forward multiples are scaled from June reports and may be stale post-WFE revisions.

14. What Must Be True

For the BULL case to be right, the following must be true:

1. AI demand sustains through 2027–28. 2027 Low-NA EUV capacity (~85 units) is absorbed; the investigated 2028 +30% add is confirmed and also absorbed; memory capex does not digest in 2028. Falsification test: two consecutive quarters of declining shipments, a 2027 guidance cut, or the 2028 capacity add being cancelled/deferred.
2. China/export regime does not materially worsen. China revenue stabilizes around the ~20% guided level; no new country-level SME ban or formal escalation strands the H2 China ramp. Falsification test: a new restriction that compresses China revenue below ~15% of sales or blocks service/replacement parts.
3. Margins march toward 56–60% by 2030. Gross margin holds the 54–56% guide in 2026, reaches ~56% in H2, and the 2030 model’s 56–60% band is delivered. Falsification test: GM stalls below 53% for a full year or pricing-power commentary reverses.
4. High-NA reaches scaled HVM on schedule. Intel’s 18A milestone broadens to more layers/products; TSMC adopts EXE for A14 or its successor; 4–5 EXE recognitions in 2026 are met and exceeded in 2027. Falsification test: repeated High-NA yield or availability slips, or TSMC indefinitely extends low-NA.

For the BEAR case to be right, the following must be true:

1. A 2026–27 digestion arrives. Customer over-build (memory +75%, TSMC $60–64B capex, synchronized HBM builds) leads to order push-outs or a flat/down 2027. Falsification test: 2026 lands at/above the top of the €43–45B range with 2027 guidance raised or confirmed above €50B.
2. The 95th-percentile multiple normalizes. Growth or sentiment cools and the trailing P/E compresses toward 30–40× regardless of earnings. Falsification test: the multiple holds in the high-40s/low-50s through a full year of in-line results.
3. A China/geopolitical shock impairs revenue. Export escalation compresses China new-tool sales and eventually the service base, or a Taiwan contingency disrupts demand. Falsification test: China revenue stabilizes at ~20% and export discussions resolve without material new curbs.

Synthesis: Bull and bear agree entirely on the business (durable monopoly) and disagree only on cycle timing, geopolitics, and the price paid. The watch-list is therefore order momentum, H2 execution, China policy, and the multiple.

15. Source Appendix

(See the dedicated Source Appendix — Appendix B in the combined report — for the full citation list. Primary sources: ASML Q2 2026 Form 6-K filed 2026-07-15, accession 000162828026048235, EX-99.1 press release, EX-99.3 U.S.-GAAP statements, EX-99.4 statutory interim report; ASML FY2025 Form 20-F filed 2026-02-25; ASML Q2 2026 earnings call transcript, 2026-07-15; ASML 2024 Investor Day 2030 model; ROIC.ai MCP tools; AZI valuation_index own-history percentiles, 2026-07-17; FactorsToday API. Third-party aggregator/AI signals are labeled as such and reconciled to filings.)

APPENDIX A — Standard Diligence Questionnaire

ASML Holding N.V. (NASDAQ: ASML) — Standard Diligence Questionnaire Appendix

Supplemental to the research memo. Grounded in primary filings (FY2025 20-F; Q2 2026 Form 6-K filed 2026-07-15; Q2 2026 earnings call). Fact / Interpretation / Assumption labeled where it matters. No price target, no buy/sell. Update of the 2026-06-09 (ASMLF) appendix; every answer refreshed for the Q2 2026 print.

General

What thoughtful questions have other investors asked about this company? The recurring, high-quality questions — updated for the Q2 2026 print — are: (1) Is the memory-led super-cycle (memory systems guided +75% in 2026) a durable wave or the most procyclical leg at its most synchronized moment — TSMC capex raised to $60–64B, Micron >$25B, SK Hynix ~$20.5B, all at once? (2) Can the raised FY2026 guide (€43–45B) actually ship, given it implies Q4 revenue of €12.9–15.9B — the largest shipment ramp in company history — plus a China re-acceleration to ~€5.9B in H2 (~2.4× the H1 run-rate) to hit the ~20% China mix? (3) With quarterly net bookings disclosure discontinued, what independently verifies demand — shipments, contract liabilities (down €1.9B in H1), and capacity commitments are all that remain? (4) Is High-NA EUV now economically adopted at scale after Intel’s 18A HVM milestone, and does TSMC’s low-NA-extension strategy (A14) cap the EXE unit trajectory? (5) At a 95.4th-percentile own-history valuation (TTM P/E 54.9×), what is the drawdown if a single quarter slips — the last two beat-and-raise prints were both sold into? Note (Interpretation): the debate among serious investors remains almost never about the moat — it is about cycle, geopolitics, execution, and price. The June 2026 Lutnick EUV-in-China allegation added a sixth question: does export-control enforcement risk escalate from rhetoric to rules?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? (Interpretation) At a cyclical high — and rising. 2026 is tracking to ~€44B (+~35% vs €32.7B in 2025) after the second guidance raise in six months (€34–39B Jan → €36–40B Apr → €43–45B Jul). Q2 2026 margins are multi-year bests (GM 54.0%, operating margin 37.1%) and Q3 is guided higher still (€11.0–12.0B, GM 55–57%). The arc is 2024 digestion → 2025 reacceleration → 2026 super-cycle (Fact on numbers; the “high” label is Interpretation).

Driven by the external environment or internal actions? Both. Externally: the AI capex wave — TSMC raised FY2026 capex to $60–64B one day after ASML’s print; the memory trio is in a synchronized ~50% HBM capacity build (Fact). Internally: ASML is the supply bottleneck and is committing to relieve it (+30% Low-NA EUV capacity for 2027 to ~85 units, +30% immersion to ~170, a further +30% for 2028 under investigation — all within existing footprint), and is now exercising explicit pricing power (CFO Dassen, 2026-07-15: “the current environment provides more flexibility for pricing… we’re executing on that”).

How stable are revenues? (Fact/Interpretation) System sales remain lumpy and cyclical (per-machine recognition; Q2 2026 units 91 vs 76 a year ago), but the Installed Base Management annuity is now 29.6% of revenue, +31.8% YoY, guided +30% for FY2026 — a larger, faster-growing cushion than at the prior report (25% of revenue). Blended revenue is more cyclical than software but less so than any WFE peer.

Outlook for products/services? Strong, and newly quantified: FY2026 EUV systems guided +>45% (~65 Low-NA shipments), non-EUV ~+25% (~130 immersion), memory-related systems +>75%, advanced-logic foundry systems +>25%. 2027 Low-NA EUV capacity is “close to being fully covered with orders” and “a significant number” of 2028 orders are already in hand (management statements — hypotheses, but now capacity-commitment-backed; Fact that they were said).

How big will this market be? Growing, and estimates were revised up across the window: Citi bull-case WFE $145B (2026) → $200B (2027) → $250B (2028); SEMI ~$139B (2026) / ~$156B (2027). ASML’s 2024 Investor Day 2030 model targets €44–60B revenue at 56–60% GM — and the FY2026 guide midpoint (€44B) already touches the model’s 2030 low end four years early (Fact), which is why the 2027-06-10 Capital Markets Day is expected to revise the model (Interpretation). International by nature: H1 2026 sales were 39.2% South Korea, 28.3% Taiwan, 15.9% China, 10.7% US.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? (Interpretation) In EUV, not at all — 100% share since the product existed; nothing in the window touches the Zeiss/Cymer/system-integration lock-in (the xLight free-electron-laser startup remains pre-commercial). The Intel High-NA HVM milestone (2026-07-15) extends the lead into the next platform generation. In DUV, stably competitive (Nikon/Canon); WFE overall remains a stable oligopoly now re-rating in unison.

How profitable is the business (ROIC, ROE)? (Fact) FY2025: ROIC 38.8%, ROE 59.9%, return on capital 47.2%, ROA 19.4% (ROIC.ai annual ratios). H1 2026 net margin 31.4% on 53.5% GM. Among the highest sustained returns in large-cap tech, with net cash on the balance sheet.

How profitable is the industry — competitors, barriers? Highly profitable oligopoly; lithography captures a disproportionate share of the WFE profit pool, and ASML’s Q2 GM (54.0%) exceeds every equipment peer’s. EUV barriers remain effectively insurmountable (proprietary tech + sole-source Zeiss optics — including a separate, non-fungible High-NA optics line — + Cymer source + €4.7B/yr R&D scale + installed-base captivity feeding the IBM annuity).

Can the business be easily understood? Yes at the “monopoly toll-collector on advanced chipmaking” level; the underlying physics is extraordinarily complex — which is itself the moat. One nuance new this quarter: with bookings disclosure discontinued, the demand verification layer of the story got harder to understand even as the business got simpler (Interpretation).

Can it be undermined by foreign low-cost labor? No. The threat vector is state-sponsored technology development (China), and EUV has resisted that for a decade-plus. Note the June 2026 escalation in tone: US Commerce Secretary Lutnick told ASML executives he is concerned an EUV machine may have reached China; ASML’s denial is credible on logistics grounds (systems are ~180 tons, tracked, serviced, remotely monitored — Fact), and no public evidence of a breach was presented.

Do brands matter? Nature of competition? Not “brand” in a consumer sense; competition is technological capability and there is no alternative at the leading edge. Customers co-develop nodes with ASML (Intel’s 18A dual-qualification on High-NA is the latest instance — Fact).

Customers’ switching costs? Extreme and rising. A fab is architected around ASML tools; the installed base locks in a decade of service/upgrade dependence — and customers are now signing long-term agreements with their own customers to justify the capacity they are ordering from ASML (management statement, hypothesis), which raises the cost of cancellation on both sides.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? (Interpretation) Yes — the EUV/High-NA technology platform, IP, customer relationships, and installed base are worth vastly more than carrying value. Book equity is €21.8B against a ~€584B market cap: ~96% of the price is franchise/growth value, i.e., there is no asset-value floor under the stock.

Off-balance-sheet liabilities? None material flagged. Standard leases; supply-chain purchase commitments. Contract liabilities (customer prepayments) total €17,428.6M — a favorable item, customers pre-funding working capital — but note they declined €1,944.0M in H1 2026 (down-payment burn), part of the H1 working-capital swing (Fact).

How conservative is the accounting? (Interpretation) High quality: US GAAP; SBC trivial (€89.6M in H1, 0.5% of sales); ETR stable ~17.3–17.5%; no one-offs besides unquantified “technology and IT transformation” costs that pushed Q2 R&D above guidance (a new line to monitor — Fact). Caveats: quarterly cash flow diverges wildly from earnings on down-payment timing (H1 2026 operating cash flow −€482.5M against €5.67B net income; Q4 2025 OCF was +€11.4B) — annual FCF is the only valid lens; the EU-IFRS/US-GAAP gap (H1 net income €5,235.7M vs €5,674.3M, development-expenditure capitalization) matters if mixing series; and the loss of the bookings series is a disclosure regression, not an accounting one.

How CapEx-hungry is the business? Still remarkably capital-light: H1 2026 capex €808.4M = 4.5% of sales. It will trend up into the 2027–2028 capacity adds (+30%/+30%), but management states these fit within the existing footprint (no new clean rooms); the new campus groundbreaking expected in 2026 is for beyond 2028. The heavy investment remains R&D (€2.46B H1, 13.6% of sales), expensed not capitalized.

Capital Allocation & Management

How much FCF, and how is it used? FY2025 FCF €11,027.3M (Fact). H1 2026 FCF was negative €1,290.9M on the receivables build (+€4.2B) and down-payment timing — while H1 capital returns were €3,733.0M (€1,655.5M dividends + €2,077.5M buybacks), i.e., H1 returns were entirely balance-sheet funded; net cash fell from ~€10.6B (YE2025) to ~€5.6B (Fact; Interpretation: the H2 cash-recovery assumption is now load-bearing, though Q4 collections historically reverse the pattern). Philosophy unchanged: fund R&D, modest capex, return the rest via growing dividends + buybacks.

Significant acquisitions recently? No. Historic deals were strategic supply-chain control (Cymer 2013, HMI 2016, Berliner Glas 2020; 24.9% Carl Zeiss SMT stake) — cost/control-based, not revenue-synergy (Greenwald-consistent).

Buying back shares? Yes — €1.0B in Q1, €1.08B in Q2 2026 (~€1.04B/quarter, ~0.8M shares at ~€1,375 avg in Q2) under the €12B 2026–2028 program (~€9.9B remaining at 2026-06-28, computed). Pace verdict (Interpretation): mechanical run-rate — on track with program mechanics, but roughly half the H1 2025 cash pace (€4.08B); the heaviest buying was done at the early-2025 trough (~$683), and the prior 2022–2025 program under-executed (€7.6B of “up to €12B”). The record reads price-insensitive rather than actively procyclical — but buying continues near all-time highs at a ~95th-percentile valuation. Share count down to 384.1M outstanding (−1.4% YoY on WA basic).

Issuing large amounts of new shares to insiders? No. SBC €89.6M H1 2026 (~0.5% of sales); employee-plan issuance (€71.7M H1 proceeds) is more than offset by buybacks; net dilution is negative.

Compensation policy / motivations of management? (Fact, FY2025 20-F Remuneration Report) AGM-approved policy: STI cash bonus max 150% of base (zero below threshold); LTI 2025–2027 weighted Relative TSR 25% + financial strategic value drivers 35% + non-financial measures 40% (including the Technology Leadership Index — management is paid partly on maintaining the technology lead, unusual and moat-relevant); vesting capped at 200%. Insider ownership ~0.8% — alignment runs through comp design, not founder stakes. Insider-transaction note: N/A by structure — ASML is a foreign private issuer exempt from Section 16; zero Form 3/4/5 exists in EDGAR 2024–2026 (Dutch AFM registers dealings domestically). No management or board changes in the window; restructuring context: ~1,700 management/coordination roles cut against ~1,400 engineering positions created (announced January 2026).

Valuation & Market Data

ADR, MLP, or K-1 issuer? Neither MLP nor K-1, and not a ratio-ADR: NASDAQ ASML is the Euronext Amsterdam ordinary share listed in the US, 1:1; OTC ASMLF is the same share’s OTC line. Dutch dividend withholding tax (15%) applies to US holders; reclaim mechanics under the US–NL treaty.

Dividend policy? Growing dividend, modest payout: FY2025 total €7.50/share (+17%; three €1.60 interims + €2.70 final paid in Q2 2026). First 2026 interim €1.88/share, payable 2026-08-05 — +17.5% vs the 2025 interims (Fact; extrapolation to a full-year ~€8.75–8.80 is Assumption). Yield ~0.5%; payout ~26–30%. Policy: “growing dividends + buybacks.”

How profitable is the business? Exceptionally, and improving with scale: Q2 2026 GM 54.0% (best since at least 2023), operating margin 37.1% (+250bps YoY), net margin 31.3%; H1 operating leverage confirmed (opex +5.5% vs sales +17.2%). FY2026 GM guided 54–56%; H2 GM ~56% per CFO.

Is net income diverging from cash from operations? Annually, no (FY2025 OCF €12.66B vs NI €9.6B; FCF ~€11.0B). H1 2026, yes — sharply: OCF −€482.5M and FCF −€1,290.9M against NI €5,674.3M, driven by AR +€4.2B (late-Q2 shipment linearity) and contract-liability burn (Fact). This is the established down-payment timing pattern (Q4 2025 OCF +€11.4B), not deterioration — but the magnitude is larger than prior years and the H2 recovery is now a load-bearing assumption; watch the Q3 print for conversion (Interpretation). Valuation context: TTM P/E 54.9×, EV/S 16.4×, EV/EBITDA ~43.5×, P/B 26.7× (filing-based, EUR/USD 1.15) — above ASML’s own 2021 bubble-era peaks; AZI own-history composite 95.4th percentile (2026-07-17; was 96.4th on 2026-06-08 — earnings are growing into the multiple).

Risks & Downside

What factors would cause the stock to decline? (1) H2 2026 execution — the raise implies Q4 of €12.9–15.9B (the largest shipment ramp ever) plus logic systems flipping from −10% H1 to +25% FY and China re-accelerating to ~€5.9B H2; a slip lands the year at/below the low end, and at the 95.4th percentile “in line with the old guide” is a miss (Interpretation on arithmetic Facts). (2) AI-capex digestion / memory air-pocket — the +75% memory driver is financed by synchronized, procyclical builds (TSMC $60–64B, Micron >$25B, SK Hynix ~$20.5B, ~50% HBM adds); the CY2023 precedent was industry memory capex −19% with individual makers −40–50% (Marathon capital-cycle framing). (3) China/export escalation — the Lutnick EUV allegation (2026-06-18/19), the Netherlands joining Pax Silica (2026-06-23), and the live Feb-2026 bipartisan letter seeking a blanket country-level SME ban — while management quietly dropped the export-control bandwidth sentence from guidance. (4) Multiple compression — TTM P/E 54.9×, no asset-value floor (equity €21.8B vs €583.7B market cap); the stock has already done −52% (2022) and −64% (lifetime max drawdown) twice in five years (Fact). (5) Customer concentration — top-4 = 61.2% of 2025 sales; TSMC alone 23.9%. (6) Taiwan geopolitical shock. (7) Sector-factor mean reversion — the +133% trailing year is factor-carried (R² 0.83; Semis factor z +2.42, Tech 63d z +3.02), so de-rating can come from the sector even if execution is flawless (Interpretation). (8) Disclosure loss — with bookings gone, the first sign of order weakness will arrive late and unquantified.

Risk of a catastrophic loss? (Interpretation) Low at the business level (monopoly, net cash ~€5.6B, 100% EUV share, capital-light). The one catastrophic tail remains a Taiwan conflict, which would devastate the entire semiconductor supply chain, not just ASML.

Chance of a total loss? Negligible barring systemic geopolitical catastrophe. The realistic adverse case is a severe drawdown, not permanent impairment: the bear scenario (AI-capex air-pocket + China escalation, trough 2028 revenue ~€36B at a 20–28× exit multiple) implies −45% to −65% peak-to-trough (Assumption-labeled scenario range, not a prediction) — consistent with the stock’s own 2022 (−52%) and lifetime (−64%) drawdown history.

Recent News & Events

Has the business environment changed recently? Yes — materially, in both directions. Positive (Fact): Q2 2026 beat (€9,326.5M sales / 54.0% GM, both above the high end of guidance) with FY2026 guidance raised a second time to €43–45B / 54–56% GM; capacity escalated (+30% Low-NA EUV to ~85 units and +30% immersion to ~170 for 2027, a further +30% for 2028 under investigation, 2027 “close to fully covered with orders”); explicit pricing power (“room to raise prices”); Intel Foundry entered HVM on 18A (Panther Lake) using High-NA EUV at NXE-matched yields — the first high-volume logic product on EXE (2026-07-15); TSMC raised FY2026 capex to $60–64B the next day; next CMD dated 2027-06-10. Negative/watch (Fact): the Lutnick EUV-in-China allegation and ASML’s denial; the Netherlands joining Pax Silica; quarterly net bookings disclosure discontinued (last print: Q4 2025 bookings €13.2B, backlog €38.8B); the export-control bandwidth sentence dropped from guidance wording; and the tape itself — two consecutive beat-and-raise prints sold into (April −6%, July +2.2% then faded to −1.7% net), i.e., good news is being absorbed, not rewarded (Interpretation).

Significant acquisitions? None.

Change in accounting policies? None flagged; continues US GAAP primary reporting (statutory interim under EU-IFRS). Notable disclosure change, not accounting: bookings/backlog no longer published quarterly; FY2026 effective tax rate (~17%) newly quantified.

Recent changes — markets, facilities, management? No management/board changes in the window. Facilities: capacity +30%/+30% within existing footprint; new campus groundbreaking expected in 2026 (post-2028 capacity). Product: NXE:3600D end-of-life declared (sold out; 1–2 units may slip into 2027); 2027 EUV mix shifts to NXE:3800E/F (~45% effective wafer-capacity uplift framing per CFO). Markets: China stepping down (36.1% of 2024 sales → 29.1% 2025 → 15.9% H1 2026 actual, ~20% FY guide) while Korea rose to 39.2% of H1 sales on the memory build.

APPENDIX B — Source Appendix

ASML Holding N.V. (NASDAQ: ASML) — Source Appendix

All figures in the memo and the diligence appendix trace to the sources below. Primary (filings) prioritized over secondary; third-party aggregator/AI signals labeled as such and reconciled to filings (filing wins on any conflict). Accessed/pulled 2026-07-18 unless noted.

Primary — SEC Filings (SEC EDGAR, CIK 0000937966)

Primary — Management Commentary (treated as hypothesis, validated vs. filings)

Secondary / Quantitative Helpers (aggregated data — not primary; reconciled to filings)

News & Media (validated articles; URLs + dates)

Analytical Frameworks

Notes on Currency & Listing Conventions

• ASML reports in EUR under US GAAP (the statutory interim report is EU-IFRS/IAS 34, unaudited — H1 2026 EU-IFRS net income €5,235.7M vs US-GAAP €5,674.3M, the gap being development-expenditure capitalization). US-GAAP series used throughout for consistency with the prior report.
• EUR/USD = 1.15 used as the conversion convention (stated Assumption; recent range ~1.15–1.16 — at 1.16 all EUR per-share figures shift by <1%). Price $1,747.58 (2026-07-17) = €1,519.6.
• NASDAQ ASML = the Euronext Amsterdam ordinary share listed in the US, 1:1 = OTC ASMLF — the same security on three lines; no ADR ratio. Dutch dividend withholding tax (15%) applies to US holders (reclaim under the US–NL treaty). Prior report was written on the ASMLF line; this update uses NASDAQ ASML — identical underlying.
• Position-agnostic: this appendix, the diligence appendix, and the memo body carry no price target and no buy/sell recommendation. Third-party price targets quoted above (sell-side/contributor) are cited as evidence of expectations, not endorsed.