ASML Holding

Refresh delta (since 2026-10-05)

What changed

1. The Q4 guide came in narrower and higher than the old range implied. Q4 2026 total net sales are guided at €14.2–15.2B at a gross margin of around 57%, with Installed Base Management (IBM) around €3.7B, R&D around €1.2B and SG&A around €0.4B. FY2026 is now €44–45B (was €43–45B) at a GM of around 56% (was 54–56%). The October dossier's implied Q4 was €12.9–15.9B with a midpoint of €14.4B. The new midpoint is €14.7B, which is about 1.5x the €9,718.1M record set in Q4'25.
2. A record Q3: €11,662.0M sales, a 57.5% GM, €4,960.4M net income. Sales landed at the high end of the €11–12B guide, and the gross margin beat the 55–57% guide. Net income includes a €509.7M gain on equity investments, which the CFO calls the revaluation of the first Mistral tranche. Operating income was €5,172.8M (44.4%).
3. The down-payment float rebuilt. US GAAP non-current contract liabilities went from €3,570.3M to €9,300.6M in the quarter (+€5.73B). Q3 operating cash flow was €11,797.6M and free cash flowFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. €11,350.4M. Nine-month CFO of €11,315.0M now covers nine-month net income of €10,634.7M. The October dossier's main forensic worry (H1 CFO −€482.5M, contract liabilities down €1.9B) has reversed. The filing does not split the €5.7B by customer or year, but the non-current classification means delivery more than a year out. That fits management's "a lot of order intake for 2028" [transcripts/2026-q3.md].
4. Supply shortfall stated in the filed release, not only on the call. The CEO statement says: "we now expect that our supply as currently planned for both 2027 and 2028 will not fully meet the demand of our customers." It also says some customers want to place orders "for 2029 and even 2030", and that ASML is "investigating ways to increase capacity for 2028 beyond the 30% increase we previously communicated". ASML gave no numeric 2027 guidance. The longer-term update is set for the Capital Markets Day on 2027-06-10 [6-k-2026-q3-presentation.md p.4].
5. China rose to 23% of system sales (Q2: 14%), and the mix shifted toward immersion DUV. By ship-to location, Q3 net system sales were Taiwan 32%, South Korea 31%, China 23%, USA 13% and rest of Asia 1%. In Q2 they were Korea 43%, Taiwan 30%, China 14%, USA 9% and Japan 4%. By technology, EUV was 49% (Q2: 57%) and ArFi 39% (Q2: 29%). Units were EUV 17, ArFi 40, ArF dry 10, KrF 35 and I-line 19 (121 in total; Q2 had 16 EUV and 23 ArFi). The CFO still guides China at about 20% of the total FY2026 business [transcripts/2026-q3.md]. Basis note: the October dossier's 15.9% was H1 total net sales by customer location from the statutory report, which is a different basis from p.9's system-sales ship-to split. The two are not directly comparable.
6. Capital return stepped up. ASML bought back about €4.1B of shares in Q3 (about 2.8M shares), against €1.08B in Q2. The interim dividend is €1.88, payable 2026-11-04 [press release; presentation p.11]. Basic weighted shares were 383.1M (Q2: 384.5M).

What held

The structural thesis is unchanged: the EUV monopoly, Zeiss single-source optics and the pre-paid order model (now visibly refilling). Lenses 1–3 carry. The Chinese immersion-DUV report flagged on 2026-10-05 got no comment in anything read here, so it is unresolved, not cleared.

Phase A — Understand the business

Company Overview — carried, see the previous dossier.

Supply Chain — carried, see the previous dossier (Zeiss optics not fungible between Low-NA and High-NA).

Competitive Advantages — carried. One addition: ASML and TSMC announced an initiative to move to 12-inch photomasks for High-NA EUV. ASML claims a ">40% productivity increase for a dual mask exposure" [presentation p.6]. The video interview says the other major customers joined, and that the reticle tools arrive "in the first half of the next decade" [transcripts/2026-q3.md].

Segments / mix

• EUV system sales ~€4.10B, ArFi ~€3.27B. In Q2 EUV was €3.8B including one High-NA [prior dossier].
• End-use: Q3 system sales were logic 54% / memory 46% (Q2 51/49). Year to date, total net sales are logic €11.0B, memory €10.3B and IBM €8.5B, against FY2025 full-year totals of €16.1B / €8.4B / €8.2B [p.10]. Memory has already passed its FY2025 total by Q3.
• FY2026 by line (CFO, video interview): EUV up "over 45%", non-EUV about 20%, IBM about 50%, China about 20% of the total [transcripts/2026-q3.md].
• High-NA: the 6-K does not disclose High-NA units or revenue for Q3, and the video interview gives no count. The 15:00 CET analyst call, where this is normally asked, was not read. What is on record: TSMC plans High-NA HVM around 2030, Samsung from 2028, and Intel runs it in HVM with more layers [transcripts/2026-q3.md, secondary].

Phase B — Measure performance

Earnings Result (Q3 2026, reported 2026-10-14) — re-run

• Sales €11,662.0M; GP €6,699.8M (57.5%, the highest in the five quarters shown); R&D €1,174.4M; SG&A €352.6M; operating income €5,172.8M (44.4%).
• Gain on equity investments €509.7M (Mistral). Excluding it, pre-tax income is about €5,241M. Tax was €925.0M (ETR 16.1%). Equity-method profit was €134.7M. NI was €4,960.4M and basic EPS €12.95 (diluted €12.93).
• Ex-Mistral EPS ≈ €11.6.
• Cash flow: CFO €11,797.6M; capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. €447.2M (PP&E €413.3M plus intangibles €33.9M); FCF €11,350.4M. €4,595.1M went into short-term investments. Buybacks were €4,071.0M, dividends €721.3M, and €1,000.4M of debt was repaid (LT debt now €1,963.0M).
• Balance sheet at 2026-09-27: cash plus short-term investments €12,735.1M; receivables €9,252.8M (Q2 €7,252.6M); inventory €11,907.3M; total assets €58,981.2M; equity €22,063.9M; equity-to-assets 37.4% (Q2 43.5%), because contract liabilities grew faster than equity.
• Guide check: sales €11.66B against the €11–12B guide was a hit (high end). GM 57.5% against 55–57% was a beat, which the CFO attributes to IBM upgrades (IBM was €3.3B against about €2.9B guided) [transcripts/2026-q3.md]. Consensus: secondary coverage reports a small revenue miss (€11.72B consensus) and a large net-income beat (€4.1B) [unverified: tikr.com / marketcapof.com, 2026-10]. Neither consensus figure is from a primary source.
• FTEs: 44,681 payroll (Q2 43,938) and 1,078 temporary (Q2 696). The 55% rise in temporary staff is consistent with the output ramp.

Earnings Calls — partial re-run

Source: ASML's own Q3 video interview, machine-transcribed by Earnings Whispers [transcripts/2026-q3.md, secondary]. The analyst call transcript was not read, so there is no analyst Q&A and no High-NA count.

• Tone: the most confident of the last five quarters. The CEO says customers "expect that their supply will not meet the demand at least until 2028", are "accelerating their plan" and are "announcing new fabs". On ASML's own capacity: "30% in 27, 30% in 28 will not allow us to meet our customer demand."
• 2026 shipments: about 65 EUV systems (>50% above 2025). Immersion is "about the same amount of tools as last year", after a plan earlier in the year for fewer.
• Pricing: the "value-based pricing" talks with key customers are complete. The effect is "not really visible in 26", appears partly in 2027 and fully from 2028.
• Other capacity levers: IBM productivity upgrades; an EUV 3800F at 230–260 wph in small volumes in 2027; immersion tools specified at 330 wph from 2027.
• New vocabulary: "orders for 2029 and even 2030", "supply … will not fully meet the demand". Still not said: no bookings or backlog figure (none published since Q4'25: €13.2B bookings, €38.8B backlog [prior dossier]).

Comps — carried, see the previous dossier. Not re-sourced.

Catalysts and evidence since 2026-10-05 — re-run

• 2026-10-14 Q3 print (record sales and GM, Q4 guide €14.2–15.2B). Secondary reports put the pre-market move at about −3% on a small revenue miss [unverified: tikr.com / marketcapof.com].
• The −13% move between 10-05 and 10-27 is not explained by anything read. No primary source or company statement covers it. It is recorded as unexplained rather than given a reason.
• 2026-11-04 interim dividend payable. 2027-01-27 Q4 print. 2027-06-10 Capital Markets Day.
• The October dossier's watch-list, scored: (1) Q4 load: confirmed and tightened. (5) China share: up (23% of system sales), and the FY "~20%" guide holds. (6) High-NA 4–5 in 2026: not testable from what was read. MATCH Act / Chinese DUV: no new primary information.

Bearing on TSMC's falsifier limb ("leading-edge wafer capacity moving into surplus"): not fired, and not directly testable from ASML. ASML sells the tools that create future capacity. It does not measure wafer utilization or pricing, which is what a surplus test needs. What ASML's Q3 does say, all from the customer side:

• Customers expect supply short of demand at least until 2028, and some want to order tools for 2029–2030 [press release; transcripts/2026-q3.md]. That points away from surplus inside the call's horizon.
• The same facts are the ingredients of a later surplus. Customers are "accelerating" plans and "announcing new fabs". ASML's own 2027 and 2028 capacity is fully committed, and +€5.7B of down payments arrived for deliveries more than a year out. Logic was 54% of Q3 system sales and Taiwan 32% (Q2: 51% and 30%) [presentation p.9]. If those orders convert, leading-edge capacity additions in 2028–2030 will be large. Whether that becomes a surplus depends on demand then, which ASML cannot show.
• Split, foundry vs memory: memory (46% of Q3 system sales, €10.3B year to date) and Korea are a large share of the order flow. Not all of the "shortage" language is about leading-edge logic. ASML does not disclose orders by customer or node.
• Verdict for the call file: no number in ASML's 6-K tests "surplus". The qualitative read is shortage through 2028, with capacity additions accelerating behind it.

Phase C — Judge people & books

Management — carried. One capital-allocation note: Q3 buybacks of €4.1B were 2.8x Q2's, made as the Euronext price ran toward its 52-week high. The CFO's €1.0B debt repayment cut LT debt to €1.96B.

Forensic Red Flags — re-run

• Pre-funding: the October flag cleared for now. Non-current contract liabilities rose +€5.73B q/q. The current portion sits inside "current liabilities" (€24,904.3M, up from €22,154.6M) and is not broken out in the summary statements, so total contract liabilities cannot be computed until the 20-F. Open: a refill this large in one quarter could be a few very large 2028 prepayments. The customer concentration behind it is not disclosed.
• Receivables still climbing: €3,023.0M (YE25) → €7,252.6M (Q2) → €9,252.8M. Annualized Q3 DSO is about 72 days, about the same as Q2's ~71. Collection on Q4's large shipments is the next check.
• Earnings quality: €509.7M of Q3 NI (about 10%) is a non-operating Mistral revaluation. Equity investments rose from €1,324.9M to €2,239.4M, including €405.7M of new purchases in Q3 (presumably further Mistral tranches; the filing does not name them). These are Level-3-style marks, and a reversal would flow through NI the same way.
• Equity-to-assets fell to 37.4%. That is a liability-mix effect (prepayments), not leverage. LT debt is €1.96B against €12.7B of cash and short-term investments.
• SEC LR/AAER: not re-fetched (the prior run on 2026-10-05 found 0/0). No new regulatory item surfaced in what was read.

Phase D — Project & stress-test

Forward Projection — FY2026 only, re-run

model.xlsx was rebuilt 2026-10-27 with Q3 in the data sheets. No valuation is read off the sheet.

FY2026 EPS bridge: 9M basic EPS €27.67 [6-K]. For Q4: sales €14.2 / 14.7 / 15.2B at 57% GM, opex €1.6B, other income €0.05B, ETR 17%, equity-method income €0.1B, about 381M shares. That gives Q4 NI of about €5.53 / 5.77 / 6.01B and Q4 EPS of about €14.5 / €15.1 / €15.8, so FY26 EPS ≈ €42.2 / €42.8 / €43.4. The October base was €39.6. About €1.1 of the gap is the Mistral gain. Euronext €1,442.4 ÷ €42.8 is about 34x FY26 base (October: ~42x). FY2027–28: not re-forecast. No numeric 2027 guide was issued. The qualitative inputs since October, all pointing the same way: supply-constrained 2027, value pricing partly in 2027, EUV 3800F and 330-wph immersion. The October base (€52B / 56.5% for FY27) is carried as-is and is not updated here.

Bull vs Bear — re-run

• What the bull side gained: both of the October dossier's checkable risks (the Q4 calendar and the working-capital drain) went the bull's way in the cash-flow statement. Supply-short language is now in a filed release, and the multiple on house FY26 EPS fell from about 42x to about 34x.
• What the bear side still has: (a) Q4 must still deliver about 1.5x the prior record quarter. (b) Bookings are still undisclosed: "orders for 2029–2030" is interest, and the release says it is not orders. (c) 10% of Q3 NI was a private-company mark. (d) China rose to 23% of system sales just as MATCH-style restrictions are pending, which increases exposure rather than reducing it. (e) Memory is about half of systems, and memory capex is the most cyclical demand ASML has.
• Pre-mortem (updated): it is 2028. The 2027–28 order book was real, but it was pulled forward by customers buying ahead of a shortage ASML itself announced. The fab wave arrived together, utilization fell, and 2029 orders were pushed out. The +€5.7B of non-current prepayments became the next cycle's cancellation fight.

Devil's Advocate — re-run, short

"Supply will not meet demand" is the sentence every equipment cycle ends on. ASML has removed the one number that would show demand peaking (bookings) and replaced it with adjectives. The prepayment refill is the best counter-evidence, and it should be watched: if non-current contract liabilities stop rising while the shortage language continues, the words have stopped being backed by cash.

Management Questions — carried, with three put first

• N1. How much of the +€5.73B in non-current contract liabilities comes from the top two customers, and what delivery years does it cover?
• N2. How many High-NA systems were recognized in Q3, and does the 2026 count still hold? (Not answered in anything read.)
• N3. China was 23% of Q3 system sales and the FY guide is about 20% of total. What share of China's Q3 shipments was immersion, and what part of the backlog would a MATCH-style ban touch?
• Remaining questions carried from the previous dossier.

Shelf & provenance summary (this run)

• Added: filings/6-k-2026-q3.md (cover, via ingest-filing), filings/6-k-2026-q3-press-release.md, -presentation.md and -us-gaap-statements.md (exhibits fetched from EDGAR and saved as text; the slide-chart percentages were read from the EDGAR slide images p.9–10), and transcripts/2026-q3.md (video interview, source_type: secondary).
• CSVs: 2026-Q3 row added to our figures (ECB Q3 average 1.1505, end 1.1355); Q3 actuals filled in and Q4/FY26 guide rows added in our figures; Q3 rows added to our figures. model.xlsx rebuilt. Side effect: our model also rewrote the shared models/INDEX.md and models/our figures.
• 20-F --limit 3: all three already on the shelf. 6-K --limit 12: only the Q3 cover was new.
• Failed reads: investor.asml.com quarterly-results returned 403, so the 6-K exhibits were used. The roic.ai transcript returned 403. The Q3 analyst-call transcript was not found or read. No High-NA Q3 count was found in any source read. The cause of the share move since 10-05 was not established.