ASML

We rate ASML an Avoid, with low confidence, at $1,658.52. Our central fair value is $1,372 a share, 17% below the price; weighting bear, central and bull cases 25/50/25 gives $1,302. On our central case the stock returns 7.1% a year, below both the 9.62% cost of equity and the 12% a year we require. The price needs every year's profit to be 21% above our central case. That means the Street's revenue path, in particular about 28% growth in systems sales in 2028: on the Street's revenue, with our margins and multiples, the stock is worth $1,738, just above the price.

How to read the cases. Each report values a company in three cases. The central case is our best estimate. The bear case is a severe but plausible bad outcome (here, China export rules tighten and AI capacity is digested from 2028). The bull case is the matching good outcome. The exit price is our estimate of the stock's value on 2031-12-31, the median of three methods: a multiple of profit that we judge fair (24x in the central case), the "long fade" (which lets returns on new capital drift down toward a mature level over time) and the "outside view" (what comparable companies trade at). Cost of equity (9.62%, in euros) is the yearly return investors require; we discount future cash to today at that rate.

What changed since the blind value. The blind value was $1,255, set before we read any price. Two outside reviewers (an audit reviewer and a Buy-case reviewer who argued the opposite side) then read the report. After review the value is $1,372 (+$117): a cash-payout error fixed (+$41), the Zeiss stake (ASML's holding in its optics supplier Carl Zeiss SMT) valued on its profit (+$12), and 2029 systems growth raised from 0% to +6% on SEMI's dated forecast (+$64). SEMI is the industry body of equipment makers. The report is partly unblinded because of exposures listed in assumptions.md: training memory of 2025 prices, the Q3 2026 presentation's buyback line (about 2.8M shares for about EUR 4.1B, an average near EUR 1,460 or about $1,640), and the peers' high recent P/Es.

ASML sells the lithography scanners that print chip patterns; it is the only maker of EUV scanners. Revenue was EUR 32.7B in 2025 and management guides EUR 44-45B for 2026 (+36%). Demand is set by AI chip and memory capacity: supply for 2027 and 2028 "will not fully meet the demand", and some customers want to order for 2029 and 2030 (Q3 release, 2026-10-14). Our central case takes systems sales up 25% in 2027 and 15% in 2028, both backed by planned capacity, then 6% in 2029 (matched to SEMI's industry forecast), 3% in 2030 and 6% in 2031, reaching EUR 74.3B of revenue in 2031 at a 58.5% gross margin. With free cash flow paid out as earned, that is $1,372 a share today. The price, $1,658.52, needs 21% more profit every year. On our central case the stock returns 7.1% a year, below the 9.62% cost of equity and the 12% hurdle. Avoid, at low confidence: our view and the Street's split on 2028-30. The 2027 outlook on 2027-01-27 and the Capital Markets Day on 2027-06-10 test it.

1. ASML is the only maker of EUV scanners, and its customers are concentrated

Who pays: chipmakers, concentrated. Four customers were each above 10% of 2025 sales, 61.2% together, the largest 23.9% (20-F 2025). By region in H1 2026: Korea EUR 7.1B, Taiwan 5.1B, China 2.9B, US 1.9B (6-K 2026-07-15, IFRS interim notes). By product in H1 2026: 29 low-NA EUV (NXE) for EUR 6.7B, 3 High-NA EUV (EXE) for 1.2B, 40 immersion DUV for 3.3B.

What it earns: the gross margin was 52.8% in 2025 and 57.5% in Q3 2026, lifted by EUV mix, volume and upgrades (CFO, Q3 2026 call). Price rises on the most critical products are agreed: some effect in 2027, full from 2028, and not to be repeated every year (CEO, Q3 call; management claim). The installed base (service and upgrades) is EUR 12.3B in our 2026, 28% of sales, and grows with the EUV fleet.

What limits it: capacity, not demand, for 2027-28. Low-NA EUV output is about 65 tools in 2026, +30% planned for 2027 and "close to fully covered with orders", and ASML is investigating capacity for 2028 beyond the +30% it had announced (Q2 call 2026-07-15; Q3 release). Customers pay down payments: non-current contract liabilities rose from EUR 3.6B (2026-06-28) to 9.3B (2026-09-27) (6-K 2026-10-14, Ex 99.3), which is why operating working capital is −20.6% of revenue. Capacity grows on the existing footprint (CEO, Q2 call), so capex stays near 4-5% of revenue.

What it does with the cash: a EUR 12B buyback programme for 2026-2028 (6-K 2026-01-28) and growing dividends; EUR 4.1B of shares bought in Q3 2026 alone. The model pays free cash flow out as earned.

Country and export risk is charged once: in the bear case's revenue (China rules tighten, systems −5% in 2028), not in the discount rate. The US Affiliates Rule and China's rare-earth export controls are both suspended to 2026-11-10 (20-F 2025).

2. Q3 2026 beat the guide, and the 2026 outlook was raised three times

Q3 2026 (6-K 2026-10-14): revenue EUR 11,662M against a EUR 11.0-12.0B guide (+1.4% on the midpoint), installed base EUR 3,289M above its guide, gross margin 57.5% against 55-57%, net income EUR 4,960M including a EUR 510M gain on the Mistral AI stake (non-operating; left out of our NOPAT). Q4 2026 guide: revenue EUR 14.2-15.2B, gross margin about 57%, R&D about EUR 1.2B, SG&A about EUR 0.4B; 2026 EUR 44-45B at about 56%.

Guidance record over eight quarters (printed in output.txt): median revenue beat on the guide midpoint +2.4% (range −1.8% to +7.2%, beats in six of eight); gross margin at or above the midpoint in all eight, median +1.5 points. The 2026 outlook was raised three times, from EUR 34-39B (2026-01-28) to 36-40B, 43-45B and 44-45B (2026-10-14).

An interim dividend of EUR 1.88 a share is payable 2026-11-04.

3. Our forecast: EUR 74.3B of revenue in 2031 is worth $1,372 a share

Model: model.py (output output.txt); assumptions and sources: assumptions.md (the blind record, unchanged). After the price: unblinded.py (output unblinded-output.txt). Euros throughout; dollars at the end. output.txt prints the blind method's $1,255 and each review step beside the revised figures.

Case-weighted (25/50/25, our assumption): $1,302.

• Cost of equity 9.62% in euros: German 10-year 3.409% (TradingView TVC:DE10Y, 2026-10-15) + peer median Blume beta 1.48 × Damodaran's implied premium 4.20% (2026-10-01). ASML's own Blume beta is 1.45 (9.48%, fair value $1,380). ASML earns and reports in euros, so the euro rate matches the cash flows. The 12% hurdle is in dollars: euro distributions and the exit are converted at rate-parity forwards (1.2189 at exit), then discounted at 12%.
• Distributions. Each year's free cash flow and interest is paid out at year end (EUR 13.9, 56.0, 61.6, 58.8, 66.0 and 73.2 a share for Q4 2026 to 2031), as ASML's buyback programme and dividends do; the share count is held flat, so buybacks are treated as if at fair value. Fair value is those payments plus the exit price, discounted at the cost of equity. The blind model kept all the cash to 2031 at 1.4% after tax, which lost value against a 9.62% discount rate (the audit reviewer's finding, section 8).
• Bridge at exit (central): today's cash (EUR 12.7B, kept) plus Mistral AI (EUR 2,239M at book) plus the Carl Zeiss SMT stake at the peers' 20.6x its 2026 profit run rate (EUR 374M, held flat; EUR 7.7B) less 3% of revenue as operating cash and EUR 3,066M of bonds, a loan and leases = EUR 17.4B, about EUR 46 a share of the EUR 1,574. Zeiss loans receivable (EUR 2,086M) are left out. Cash earns 1.7% before tax, debt costs 2.7% (2025 actuals, 20-F).
• Stock pay is an expense inside R&D and SG&A (EUR 202M in 2025); the share count is flat at 381.7M.
• Methods. The judged 24x on 2032 NOPAT is the peers' through-cycle 20.6x plus about 3-4 turns for the EUV monopoly and the service annuity; the value-driver formula gives 24.0x at 6% growth and a 40% return (a match we note, not support). The long fade holds a 40% return on new capital, against ASML's own roughly 85% in 2025: a harsh input by choice. The outside view is the peers' median ten-year trailing P/E, 20.6x, on 2031 NOPAT plus exit net cash. In the central case the outside view is the median, so at the peers' 20.6x instead of 24x the value is unchanged; it falls only below about 20x (18x gives $1,294).
• Consistency check. 24x implies 6.0% growth at a 40% return; the path's NOPAT grows 7.2% in 2031 and 6% in 2032. The path's return on new capital is 186% over 2027-31 because down payments fund the growth (our report checklist marks this as a "smell", a pattern worth a second look). The 2029 gross-margin dip (57.5%, from 58.5%) was set blind for a flat year and is kept with 6% growth: conservative by about the Buy-case reviewer's margin finding.

Q4 2026 (release 2027-01-27), our forecast: revenue EUR 14,878M (range 14,400-15,300), the guide midpoint plus half the median beat; gross margin 57.5% (56.8-58.3%); diluted EPS EUR 15.52 (14.70-16.31, the revenue and margin ends with the same costs). FY2026 revenue EUR 44,633M. None of the review changes touch Q4 2026.

4. The price needs every year's profit 21% above our central case

At $1,658.52 (EUR 1,479 at spot) the market value is about $633B on 381.7M shares. On FMP's mean EPS that is 26.7x 2027 (EUR 55.46, 22 analysts) and 20.1x 2028 (EUR 73.73, 25). TradingView shows a trailing P/E of 40.8.

FMP analyst-estimates, fetched 2026-10-15 after the blind push (data/analyst-estimates.json). Our 2029 and 2030 are still below the lowest single Street estimate.

Reverse underwrite (unblinded-output.txt): the price needs every year's NOPAT ×1.212 of our central (2031 NOPAT EUR 34.3B against 28.3B), and ×1.244 to earn the 12% hurdle. At our central margins that is about EUR 90B of 2031 revenue, close to the Street's EUR 91.3B path.

Price reaction (Robinhood daily closes, data/daily-closes.json): results came before the US open on 2026-10-14. ASML closed $1,717.76 on 10-13 and $1,650.00 on 10-14 (−3.9%); the VanEck semiconductor ETF SMH fell from $588.12 to $564.74 (−4.0%) the same day, so the move was sector-wide, not a verdict on the quarter. Over the prior week ASML fell from $1,804.96 (10-07) to $1,717.76 (−4.8%) and SMH from $625.03 to $588.12 (−5.9%).

5. We match the Street on 2026-27 and differ on 2028-30

Same 2026-27, different 2028-30. Our 2027 is 4% under the Street, inside its range. From 2028 we part: on the Street's revenue, systems grow about 28%, 14% and 7% in 2028-30; ours grow 15%, 6% and 3%.

Evidence on each side:

• SEMI 300mm Fab Outlook, 2026-10-13 (data/semi-2026-10-13-300mm-fab-outlook.txt): 300mm fab equipment spending $149B in 2026, growing at double-digit rates each year through 2029 to $227B (×1.52). The equipment makers' trade body, independent of ASML's price. Published before the price was read, but this analyst read it on 2026-10-15 after the price. Matched to our 2027-28, it gives ASML systems +6% in 2029, which the central now uses (+$64).
• SEMI, 2026-07-14: wafer fab equipment +21.8% in 2027 and +14.1% in 2028, close to our ASML systems +25% and +15%. It is the only dated, independent 2028 figure we found, and it supports our 2028 over the Street's +28%.
• ASML itself (6-K 2026-10-14, read before the blind commit): supply short of demand in 2027 and 2028, customers asking to order for 2029-30, capacity beyond +30% for 2028 under investigation, not committed.
• Our judgment from ASML's own record: growth slowed sharply after each earlier capacity wave (2019 +8% after +22%; 2024 +2.6% after +30%). 2030 +3% rests on this alone; SEMI says nothing about 2030.

On the Street's revenue for 2027-30 with our margins, costs and multiples, the value is $1,738 and the purchase price at 12% $1,693, both just above the price; on the Street's 2028 alone, then our growth, $1,590. So the price is the Street's revenue path valued on our margins and multiples: the market is not paying for higher margins or a higher multiple than ours, only for growth near 28% in 2028 that keeps going through 2030.

What settles it: the 2027 outlook (2027-01-27), whether the 2028 capacity step beyond +30% is committed and covered with orders, the down-payment balance through 2027, and the 2030 targets at the Capital Markets Day (2027-06-10; the 2024 targets were EUR 44-60B for 2030, whose low end 2026 has already reached).

6. The bear case ($576), and why the Street could be right

Bear ($576). China export rules tighten and AI capacity digests from 2028: systems +15%, −5%, −25%, +5%, +8%; installed base flat in 2029; gross margin to 51% in 2029 (2024's 51.3% less a point); down payments unwind to zero working capital by 2030; 16x at exit. Growth, margin and multiple all fall together (our report checklist marks this as a "smell", a pattern worth a second look). The bear's outside view (EUR 751) is 22% above its median.

Why the Street could be right (Buy). ASML has beaten its own revenue guide in six of eight quarters, raised 2026 three times, and says demand exceeds planned supply for two years with customers asking for 2029-30 slots. Down payments rose 2.6 times in one quarter. Price rises take full effect in 2028. On the Street's revenue the value is $1,738 even at our margins and multiples, and the purchase price ($1,693) clears the price. The Buy-case reviewer argued it (section 8): 2028 systems +25%, 2029 +10%, 2030 +6%, a 60% gross margin and down payments held at −20% give $1,617 on our revised model, a purchase price of $1,576 and a 10.8% return from today's price; that reviewer itself concluded Buy fails at 12% and put its path at about 20% likely.

Why we still say Avoid. On our central case the stock returns 7.1% a year, under the 9.62% cost of equity, and the price is 17% above our value. Only the Street's full 2027-30 path clears the 12% hurdle, and only by 2%; the Buy-case reviewer's strongest path does not. The one dated, independent 2028 figure (SEMI, +14.1%) matches our +15%, not the Street's +28%. We considered Hold: the case for it is that the Street path is above the price and that the 2028 capacity step could be committed in January. We stay at Avoid because both conditions are future and the price already pays for them.

7. What to watch: dates and conditions that would change our view

Catalysts:

• 2026-11-10: the one-year suspensions of the US Affiliates Rule and of China's rare-earth export controls end (20-F 2025); renewal or not.
• 2027-01-27: Q4 2026 results and the 2027 outlook (company calendar, 6-K 2026-07-15): revenue against our EUR 14,878M; 2027 against our EUR 54.6B and the Street's 57.0B.
• 2027-06-10: Capital Markets Day, updated longer-term targets (company-announced).

Falsifiers:

• A 2027 outlook at or above EUR 57.7B (our bull's 2027), or the 2028 capacity step above +30% committed and covered with orders: our 2028 is too low; re-underwrite toward the Street ($1,590 on its 2028 alone, $1,738 on 2027-30), likely to Hold or Buy.
• A 2030 revenue target at the Capital Markets Day with a low end at or above EUR 75B: our 2029-30 is too low; move it up.
• Non-current contract liabilities (down payments) below EUR 6B in any 2027 quarter, from EUR 9.3B: orders are slowing; weight the bear ($576).
• 2027 gross margin guided below 55%: price and mix are not landing; NOPAT below central.
• A US or Dutch rule after 2026-11-10 that stops immersion DUV sales or servicing in China: the bear's export lever.

8. Review: the questions our report check raised, the adversarial review and our checklist

Questions our report check raised that apply:

• Net move from the blind value (medium): net move from the blind $1,255 to $1,372, +$117 (+9.3%). Largest change: 2029 systems 0% → +6%, +$64, on SEMI's 2026-10-13 forecast (an industry association, not a management claim or a Street figure; read after the price). The other two changes fix model errors on filings read before the blind commit (+$41, +$12).
• Bull-method gap (low): the bull's judged-multiple method (EUR 3,398) is 56% above the bull median (EUR 2,177): 30x on NOPAT still growing 8%, against a through-cycle 20.6x. The audit reviewer cleared it: the bull's 30x sits below its own formula's 56x and the outside view sets the bull median.

Considered and cleared:

• Blind view against final view: blind view Avoid, final view Avoid. The blind commit stated no view in words, but its assumptions.md set the $1,255 value 23% below the Q3 buyback average (about $1,640) it had seen.
• Value against price: −17.3%.
• Gap to the Street: central 2029-30 is now 16.6% and 18.7% below the Street mean, under the 20% line, with a dated source (SEMI) for 2029; still below the lowest single Street estimate in both years, disclosed in section 4.
• Multiple and beta: the 24x exit multiple is above the peers' 20.6x, but the outside view at 20.6x is the central median, so 20.6x changes nothing. The peer beta (1.48) is harsher than ASML's own (1.45). The audit reviewer found the 20.6x not lenient (about 19.4x forward, fitting about 5% growth).
• Purchase price against beta: purchase price $1,337 at either beta.
• Share count: 381.7M = 384.1M issued (2026-06-28) − 2.8M bought in Q3 + 0.4M dilutive awards; the audit reviewer confirmed against EDGAR.
• Filings read: 20-F 2025, the H1 2026 interim report and the Q2 2024 to Q3 2026 6-K releases read.
• Cost of equity against the hurdle: 9.62% < 12%.
• Size of the review change: the reviewers moved the central value 9.3%, under 20%.
• Bridge items with filing lines: down payments, both stakes, bonds, loan and leases each have a filing line, confirmed by the audit reviewer; Zeiss loans left out (Audit 3 below).
• Batch skew is left to Capfolio's coordinating agent to judge across the day's reports.

Adversarial review (2026-10-15)

Capfolio's coordinating agent started both outside reviewers (an audit reviewer and an opposite-side reviewer arguing Buy, the "Buy-case reviewer"); the analyst writing this report could not. Each effect is measured alone on the revised central in unblinded-output.txt ("Review findings"), except the three fixes, which output.txt prints step by step from the blind $1,255. When each source was read: the 6-K of 2026-10-14 (results, statements, presentation, and the 6-K of 2026-01-28 on the buyback) before the blind commit; SEMI 2026-10-13 and 2026-07-14 after the price, on 2026-10-15.

Net: $1,255 → $1,372 (+$117), purchase $1,219 → $1,337, bear $538 → $576, bull $1,820 → $1,887, exit $2,201 → $1,919 (the exit now excludes EUR 330 a share paid out before it). View unchanged: Avoid. No second round: no fix changed the call.

Our report checklist (research-eval), run by the analyst on this draft (not independent). Its codes group the checks: B for the arithmetic and the bridge, C for the cases, S for sources, I for the integrity of the blind step, D for the decision block. In its output, "auditor" is the audit reviewer and "advocate" the Buy-case reviewer.

B1 PASS  cash earns 1.7%, debt costs 2.7%, after 17% tax (output: net interest lines)
B2 PASS  exit multiple and outside view price NOPAT; interest accrues to cash once
B3 PASS  bonds, loan, leases, operating cash, Mistral at book, Zeiss on its profit;
         no minority interest; Zeiss loans left out (conservative)
B4 PASS  stock pay expensed (US GAAP, inside R&D and SG&A), share count flat
B5 PASS  free cash flow and interest paid out as earned; not also kept in the exit
B6 PASS  EV/NOPAT on 2032 NOPAT at end-2031; trailing P/E on 2031 NOPAT
C1 PASS  each case has its own growth, margin, capex, working capital, multiple
C2 PASS  all three cases: revenue and margin from our drivers; capex bends with
         revenue
C3 PASS  (a) 24x implies 6.0% vs 6% path; (b) fade return 40% against ~85% in
         2025, named; (c) see M3; (d) bull method gap named; (e) fade return
         not above the path
C4 PASS  $1,302 at 25/50/25, labelled ours
S1 PASS  2029 tied to SEMI; 2030-31 labelled our judgment; multiple, beta, rates sourced
S2 PASS  stock pay, debt, leases, capex from the 20-F and 6-Ks (FMP stock pay
         and debt wrong for this foreign filer)
S3 PASS  FMP years are calendar years, ASML's fiscal year; euros throughout
S4 PASS  Robinhood targets dated (updated 2026-10-14 18:39 UTC)
S5 PASS  every price move gives both dates and closes; SMH checked each day
I1 FAIL  assumptions.md names the Q3 buyback average (about EUR 1,460) in its
         exposure list; seen before the model; disclosed
I2 FAIL  model.py changed after b50c95c for the three review fixes; the blind
         method's $1,255 is still printed in output.txt; disclosed above
D1 PASS  $1,372 / $1,919 / $1,337 distinct; 9.62% and 12% stated
D2 PASS  "Market expects" line and section 4
D3 PASS  Alpha rank 143 of 713, 2026-10-15
D4 PASS  dated catalysts; falsifiers with levels
D5 PASS  every auditor and advocate finding answered in the table above
D6 PASS  management claims, Street figures, SEMI forecast and our judgments labelled
SMELL M2 path return on new capital 186% (central), funded by down payments
SMELL M3 bear: lower growth, lower margin and lower multiple together

Our automated report check (scripts/research-check.py) on this draft, 2026-10-15:

PASS record: loads as the site loads it
PASS rerun model.py: matches output.txt
PASS rerun unblinded.py: matches unblinded-output.txt
PASS record fair_value: 1372 printed by the model and in the report
PASS blind commit: b50c95c: 32 files, no report, no price
PASS decision block: all labels present
FLAG 1 medium net move from the blind value 1255 to 1372: +117 dollars, +9.3%; name the largest change; high if it rests only on a management claim or Street figure found after the price
depth: research-eval, the model auditor and the advocate aimed at the flags

Both reviewers already ran; the net-move flag is answered above (largest change: 2029 on SEMI, +$64).

Blocking: none known.

9. What slowed the work (friction log)

• The filing mask missed the Q3 presentation's buyback line (shares bought for a total amount), which gives an average price; it was read before the model was built and is disclosed. The pattern now covers it.
• The blind model kept all free cash flow to the exit. For a company that generates EUR 20B+ a year and pays it out, cash earning 1.4% after tax and discounted at 9.62% lost $41 a share. The audit reviewer found it.
• FMP errors for this foreign filer: stock pay 2025 shown as 0 (filing EUR 202.3M); total debt at Q3 2026 shown as EUR 1,963M, missing the bond due in 2027 and leases. The model uses the filings.
• The worktree guard refuses compound shell commands and $VAR expansions; EDGAR's identity was read inside Python.
• Bigdata.com: three calls, all after the blind push: one search, one fetch refused for a wrong argument name, one fetch of the SEMI release. No tearsheet was used.
• The bull case's path return on new capital prints "nan%" with working capital, because its working capital falls every year and the denominator turns negative; a print quirk, not a value input.

10. Sources (read 2026-10-15)

• SEC EDGAR: 20-F 2025 (0001628280-26-011378, excerpts in data/20f-2025-excerpts.txt); 6-K 2026-01-28 (Q4 2025 release, buyback programme); 6-K 2026-07-15 (Q2 release and H1 2026 interim report); 6-K 2026-10-14 (0001628280-26-065813: Q3 release, presentation, statements); quarterly 6-K releases Q2 2024 to Q1 2026. Saved in data/; masked for price figures.
• Calls: Q2 and Q3 2026 transcripts (TradingView), masked (data/q2-26-call-transcript-masked.txt, data/q3-26-call-transcript-masked.txt).
• SEMI press release 2026-10-13 via Bigdata.com (data/semi-2026-10-13-300mm-fab-outlook.txt), with SEMI 2026-07-14, Applied Materials 2026-10-13 and BofA 2026-06-26 and 2026-09-16 from the same search; SEMI's two forecasts are used, the others recorded only.
• FMP: statements, segments, peer ratios, ASML analyst-estimates; daily prices for weekly-return betas (saved as log returns).
• Robinhood: last trade, daily closes (ASML, SMH), analyst ratings.
• TradingView: German and US 10-year yields, EUR/USD, forecasts.
• Damodaran, implied equity risk premium, 2026-10-01.
• Alpha: alpha rank for ASML, 2026-10-15 (data/alpha-rank.json).