ASMLASML Holding N.V.

ASML — dossier, 2026-09-21

Slot names-ai-capex-3. Model-drafted unattended by the cloud research routine; nothing here is human-verified. Sources and locators are in relationships.yaml; figures below carry theirs inline. Anything from background knowledge is marked [background].

Price reference: $1,714.68 close, 2026-09-21 (ROIC.ai get_latest_stock_price NASDAQ:ASML; volume 888,401 shares, so roughly $1.52bn traded that day). EUR/USD 1.14673 on the same date (ROIC.ai get_latest_forex_price FX:EURUSD) — every conversion below uses that one rate and is marked as a conversion. ASML reports in euro; the US listing is one ordinary share per ADS [background].

Shares outstanding 385,417,665 at 2025-12-31 (20-F XBRL CommonStockSharesOutstanding) → market capitalisation ~$660.9bn / ~€576.3bn.

This is a foreign private issuer. Under the skill's §3 table that means: 20-F Item 3.D risk factors and Item 5 in place of a 10-K, the latest 6-K results in place of a 10-Q, proxy skipped (not filed) and insider pattern skipped (no Form 4). The EDGAR history is 25 20-Fs and 361 6-Ks; there is no 10-K, no DEF 14A and no Form 4 to read.

S1 — Source

watchlist (owner, Apple Stocks, 2026-09-20), cluster ai_semis, coverage sec_fpi, cik 937966. No Tier 0 promotion.

S2 — Kill test

All figures in EUR from the 20-F XBRL company facts (data.sec.gov/api/xbrl/companyfacts, CIK 0000937966).

Quarterly, the two most recent (Q2 2026 6-K press release, 2026-07-15):

Backlog: remaining performance obligations €46.5bn at 2025-12-31, against €43.3bn (2024), €45.0bn (2023), €45.4bn (2022), €28.9bn (2021), €15.1bn (2020). After three flat years the backlog turned up again in 2025 — 1.4x the year's revenue.

Net debt: not computed. Cash and short-term investments were €7,582m at 2026-06-30 (€12,916m of cash alone at 2025-12-31, before €5.95bn of 2025 buybacks worked through). ASML's outstanding eurobond debt was not read this run — the 20-F's financing appendix was not opened — so no net-debt or net-debt/EBITDA figure is stated here rather than guessed. Given €9.6bn of annual net income and a maturity profile that is not a constraint on any reading of the cash flows, this is a gap in completeness, not in the survival question.

Share count over five years — shrinking, which is rare in this cluster. 419.1m diluted weighted average (2020) → 388.9m (2025), −7.2%; shares outstanding 416.5m → 385.4m. €5.95bn repurchased in 2025 and €1.1bn in Q2 2026 under a 2026–2028 programme, plus a €1.88 interim 2026 dividend paid 2026-08-05. Of the three names in this slot, ASML is the only one where the count goes down.

Analyst coverage: not obtained (no transcript was reachable — see S3).

Verdict — survives outright, and is the highest-quality balance sheet in this slot. A 52.8% gross margin and a 29.4% net margin on €32.7bn, six straight profitable years through a semiconductor downturn, €46.5bn of backlog and a shrinking share count. Nothing here threatens solvency. The kill question for ASML is not financial — it is political: 29.1% of sales go to a country whose access to the product is decided by three governments, none of them ASML's customer.

S3 — Filings, gate-minimum (FPI variant)

Read this run: the FY2025 20-F (Item 3.D risk factors in full heading form and four risk factors in body, Item 5 financial performance and MD&A, geographic and customer concentration), the Q2 2026 6-K press release (results, guidance, capacity, buyback), and the investor-relations presentation filed with it. Not read: any transcript. Equibles has no speaker-labelled transcript on file for ASML FY2026 Q2, and the fool.com fallback URL for that date returned 404. The gate-minimum spec asks for two transcripts; zero were obtained, and every forward statement below therefore comes from the written 6-K rather than from Q&A. That is this name's thin evidence.

The three biggest risks management itself discloses

1. Export controls decide what can ship, and ASML does not control the process. "under Dutch, US and other applicable laws, we are required to secure export licenses for EUV systems, specific DUV immersion systems, and some of our other products." The same risk factor adds that US measures impose "license requirements for the sale or transfer of US-origin items, as well as limitations on support by US persons for non-US origin items destined for advanced-node fabs in China" and that ASML does "not control the licensing process or approval criteria." One specific, dated exposure sits inside the next twelve months: the US 50%-ownership affiliate rule "has been suspended for one year, until November 10, 2026, as part of a trade agreement with China." A suspension, not a repeal.

1. Customer concentration, and it got worse in 2025. "Total net sales to our largest customer amounted to €7,796.7 million, or 23.9% of total net sales in 2025, compared with €4,682.4 million, or 16.6% of total net sales in 2024. In 2025, 38.0% of total net sales were made to our two largest customers." Seven points of concentration added in a single year. ASML does not name the customers.

1. Cyclicality and geographic exposure. The risk factor list opens with "The semiconductor industry can be cyclical, and we may be adversely affected by any downturn", and the international-operations factor puts a number on where the cycle sits: "Customers in China represented 29.1% of our 2025 total net sales and 36.1% of our 2024 net sales." Competition is named concretely — "We compete primarily with Canon and Nikon in respect of DUV systems" — and separately against Applied Materials and KLA-Tencor, which is worth recording because AMAT is in the same watchlist cluster and the 2026-09-20 cluster file places it purely as a supplier to the chain, not as a competitor to another member.

Latest quarter and guidance

Q2 2026 came in above guidance on both lines — €9.3bn of sales and 54.0% gross margin — "driven primarily by higher than expected Installed Base Management sales", i.e. service and field upgrades rather than new systems. Guidance:

• Q3 2026: €11.0–12.0bn of net sales, 55–57% gross margin, R&D ~€1.2bn, SG&A ~€0.4bn.
• FY2026 raised to €43–45bn of net sales, 54–56% gross margin. Against €32.7bn in 2025 that is +32% to +38%.
• Next Capital Markets Day 2027-06-10, where the longer-term view will be updated.

The most important sentence in the filing is about capacity, not revenue. "we are planning to add 30% to our 2026 low NA EUV capacity of around 65 for 2027, and we are investigating to increase capacity with another 30% for 2028. Similarly, we plan to add 30% to our 2026 DUV immersion capacity of around 130 for 2027." The 2026-09-20 cluster pass argued that TSMC's capacity discipline is the best single bubble indicator for this complex, and flagged that TSMC is absent from the watchlist. This is the same indicator one layer up the chain, published quarterly, from a name that is on the watchlist. The monopolist at the bottleneck is choosing to remove the bottleneck.

Proxy, insider pattern

Skipped — not filed. FPIs file no DEF 14A and their insiders file no Form 4, so there is no insider-ownership percentage, no compensation structure and no insider transaction pattern for this name. This is a structural gap in coverage for every sec_fpi row on the watchlist, not a gap in this run. Dutch remuneration and shareholding disclosure exists in the annual report; it was not extracted.

Red-flag sweep

• Auditor change / restatement: nothing in the EDGAR index (no 20-F/A, no NT 20-F). The efts.sec.gov full-text sweep was not run (time).
• Going concern: not applicable on these numbers; not checked with the tool (Equibles budget).
• Executive changes: not checked.
• Customer concentration: read from the 20-F directly, quoted above.
• Unresolved staff comments: Item 4A is "Not applicable" — the SEC has no open comments on this issuer. ASML's last CORRESP exchange was 2020.

S5 — Ownership

⚠️ 3spread does not cover this name. threespread.py holders ASML (CUSIP N07059202) returned 2 filers and 1,185 shares, against Equibles' 2,504 institutions and 86.0m shares. The skill routes all of S5 through 3spread; for ADR and other foreign CUSIPs it answers near-empty rather than erroring. Slot names-ai-capex-2 found exactly this for ARM and OMAB. Three confirmations now — this should be treated as a known tool limitation, not re-discovered a fourth time.

Top institutional holders, 13F report date 2026-06-30 (Equibles GetTopHolders; 2,595 rows across 2,504 institutions, 86,026,133 shares, $171.1bn):

Note that the 86.0m institutionally reported shares are 22% of the 385.4m outstanding — the rest sits in Amsterdam, outside 13F reporting, so this table is the US slice and not the register.

Two things stand out. Fisher Asset Management at number two is an independent RIA rather than an index complex — an active, concentrated holder near the top of a €576bn company. And row 10 is a put, not a holding: Susquehanna carries a 1,836,900-share notional bearish options position, the only such line in the top twelve.

13D/G: three filings in the whole history, all 13G/A, newest 2025-04-23. No 13D, no activist.

Insider ownership: not available (FPI — see S3).

Short interest (Equibles, FINRA): 1,284,921 (2026-07-15) → 1,068,098 → 1,352,364 → 1,524,269 (2026-08-31), days to cover 1.0–1.4. Against 385.4m shares that is 0.4% of shares outstanding — effectively nobody is short the ADR. The tool's 2026-09-15 estimate (~1.60m) is a model output, not FINRA data. The organised bearish expression here is Susquehanna's put line, not stock borrow.

Shelf, ATM, buyback: the question is inverted for this name. No shelf or ATM (an FPI issues under Dutch corporate authority, not an S-3 [background]). Capital flows the other way: €5.95bn repurchased in 2025, ~€1.1bn in Q2 2026 under the 2026–2028 programme, plus an interim 2026 dividend of €1.88 per ordinary share paid 2026-08-05. Nothing hangs over the stock.

Who is on the other side: large active managers, an index tail, and one options desk with a put. There is no crowded short and no activist. What is on the other side of this trade is a government licence and a multiple.

S8 — Valuation and sizing (proposed only; owner-only items OPEN)

At $1,714.68 with 385.4m shares: market capitalisation ~$660.9bn / ~€576.3bn (converted at EUR/USD 1.14673).

Against FY2025: 17.6x EV-free price/sales, 60.0x net income (€576.3bn / €9,609.4m). Basic EPS was €24.73, or $28.36 converted, so the trailing P/E on the traded ADR is 60.5x.

Out-year model. FY2026 is management's own guide; FY2027–28 are built from the capacity additions management has announced (+30% low-NA EUV and DUV immersion for 2027, a further 30% "investigating" for 2028) rather than from a growth rate.

The base case is ~$1,780 against $1,714.68 — about 4% of upside over three years, and it already assumes the FY2026 guide is met, the 2027 capacity addition is filled, and the margin keeps climbing. That is the same shape slot names-ai-capex-2 found at MRVL and ARM: a correct operating forecast that does not produce a return at today's price. Three names in two slots now model at or below the traded price.

⏳ Buy price and position size are owner-only and stay OPEN (user). Liquidity is not a constraint (~$1.52bn traded on 2026-09-21 on the US line alone, with the Amsterdam line on top).

S9 — Monitoring

Next earnings: Q3 2026 results, expected around 2026-10-14 — ASML reports on a strict mid-January / mid-April / mid-July / mid-October cadence (Q2 filed 2026-07-15, Q1 2026-04-15), derived from the 6-K cadence rather than a calendar feed. Capital Markets Day 2027-06-10, which the company itself names as the date the long-term view gets rewritten, is the larger event.

Filing feed: the Tier 0 EDGAR sweep covers cik 937966. For an FPI the feed is 6-Ks, and ASML files a great many of them (361 in the index) — the quarterly results 6-K is the one that matters and it is distinguishable by document name (form6-kquarterlyfilings.htm).

The leading indicator: announced capacity, not orders. ASML no longer publishes a bookings number in the quarterly release — the Q2 press release says only that "Our order intake remained extremely strong in the first half of the year" with no figure. What it does publish is planned capacity: ~65 low-NA EUV units in 2026, +30% for 2027, a further 30% "investigating" for 2028, and the same steps on DUV immersion off ~130 units. Whether the 2028 "investigating" becomes "planning" is the single most informative disclosure this company will make, because ASML adds capacity only against customer commitments and it is the one firm in the chain with no competitor to lose share to. A retreat from that second 30% would be the earliest credible signal that the AI build is topping — and it would come from the layer above TSMC, which the watchlist does not carry. Secondary: the China revenue share (36.1% → 29.1%) and the 2026-11-10 expiry of the suspended US 50%-ownership rule.

Trigger that converts this to a buy: the 2028 capacity step confirmed as planned rather than investigated, while the shares trade below ~28x the forward EPS on this model — roughly $1,140 on the FY2027E line. On the evidence in hand the business is not the question; the entry price is. The specific price and the size remain the owner's.