What we said in September — and what happened
The business beat the plan; the price fell through our level with its sector; five numbers were wrong
On 28 September we published ASML at $1,770.32. We called it the purest monopoly on our board, a wonderful business at a full but fair price, and gave readers two instructions: accumulate slowly, and keep some powder dry for the next cyclical scare. If that scare offered ASML near $1,365, about 22.5 times the earnings analysts expected for 2027, we said to buy with both hands. We named the risks: a pause in AI spending, a sharper cut to sales in China, and the politics of export controls. Eight weeks later, on 20 November, the shares closed at $1,266.22, 28.5% lower and 7.2% below our level. Here is the account.
| What we said on 28 Sep | What happened by 20 Nov | Grade |
|---|---|---|
| 'Accumulate slowly' | A buyer at $1,770 is 28.5% down. The chip basket (SOXX) fell 34% over the same weeks. | ✗ Early |
| 'Buy with both hands near ~$1,365' | Reached on 5 November ($1,356.65), again on 19 and 20 November. The price is now 7.2% below it. | ◆ Reached |
| 2026 guide €43–45bn; Q3 guide €11–12bn at 55–57% | Q3 €11.66bn at 57.5%, above the margin guide; the year raised to €44–45bn. | ✓ Beaten |
| Capacity +30% in 2027; a further +30% for 2028 under study | Supply 'as currently planned for both 2027 and 2028 will not fully meet the demand'; ASML is looking at more for 2028; some customers want slots in 2029–30. | ✓ Stronger |
| China to fall to about 20% of sales in 2026 | Still ~20% for the year (finance chief, as reported); 23% of system sales in Q3, from 14% in Q2. | ✓ On guide |
| 'A pause in AI spending or a sharper cut to China could take a third off the shares' | The shares fell 32% from the 2 October high, but neither happened: no customer cut its spending and China rose. | ◆ Size right, cause wrong |
| MATCH Act proposed, not enacted | Still not enacted; no floor vote found. The Dutch government has objected to it (as reported). | ✓ Right |
| ~29× 2027 earnings | 20.5×: 2027 estimates rose about 5% while the price fell. | ✓ Cheaper |
Our errors. Five numbers in the September report were wrong, and we correct them here. None changes the verdict; two of them make the business look slightly less comfortable than we said. First, we wrote that ASML raised its guidance "after second-quarter bookings of about €5.5 billion". ASML has not disclosed quarterly bookings in 2026; the finance chief said on the July call that the company is "not sharing order intakes". The €5.5bn was the figure for the second quarter of 2025. Second, we wrote that China bought "a third of ASML's sales" in 2025. China was 33% of system sales, the machines alone; of total sales, including service, it was 29.1% (€9.52bn of €32.67bn, according to the 20-F). Since ASML's forecast of about 20% for 2026 is a share of total sales, the like-for-like fall is from 29% to about 20%, not from a third. Third, we gave net cash of about €10bn. That was our feed's figure, which leaves out short-term debt. On the latest balance sheet we had, at the end of June, net cash was about €3.9bn; at the end of 2025 about €8.9bn. (It is about €10bn again today, for a reason we explain in Part IV.) Fourth, 2025 free cash flow was €11.0bn on ASML's own definition, not €11.1bn; our feed deducts spending on equipment but not on intangible assets. Fifth, we wrote that management was "close to having all the orders it needs for 2027". The finance chief said ASML was close to fully covered for its low-NA EUV machines, not for everything it sells.
Our postscript of 17 November repeated the second error ("about a third in 2025") and told readers to buy with both hands at $1,383.11, 1.3% above the level we had set; a buyer that day is 8.5% down. Everything else we re-checked was right: the 2025 sales, profit and earnings per share, the guidance history, the second-quarter figures, the capacity plan, the €7.50 dividend and the €1.88 interim, the €12bn buyback programme, the names and roles of management, the date of the annual meeting and the arithmetic of our multiples.
★ Down 28.5% — and the chip basket down 34%
Eight weeks in which the news from ASML was good and the price fell with everything around it
| Date | What happened | ASML that day |
|---|---|---|
| 2 Oct | The high since our X-Ray. | $1,867.31 |
| 14 Oct | Q3 results: above the margin guide, the year raised. Some reports said sales slightly missed analysts' estimates; the chip basket fell 4.5% the same day. | −3.9% |
| 22 Oct | A retreat across chip shares after Meta raised its spending plans and its own shares fell (as reported). | −6.9% |
| 30 Oct | Intel fell 29% on a weak fourth-quarter forecast; Alphabet had fallen 13% the day before (as reported). Chip basket −8.2%. | −6.6% |
| 5 Nov | The semiconductor index fell 12%, its second-worst day on record; reports cited fears of AI regulation after the US midterm results, hedge funds cutting positions and a slowdown in spending. ASML bought 125,000 of its own shares that day. | −9.3%, $1,356.65 |
| 16 Nov | The week of turmoil at Nvidia's top; a new US rule on access to AI computing abroad; the 10-year Treasury yield at 5.15% (as reported). | −5.9% |
| 20 Nov | The Nasdaq's worst week since March 2020; Nvidia −14% after a Reuters report of an SEC inquiry, which is a report, not a finding (CNBC). Chip basket −10.6%. No ASML news. | −6.6%, $1,266.22 |
Nothing that happened was about ASML. We read every release ASML published in the period and found no warning, no change of management, no lost order and no new rule aimed at the company. On its own results day, the only day of the seven that concerned it directly, ASML fell less than the chip basket. On the three worst days for the sector (5 November, 20 November and 30 October) it fell 9.3% against 12.2%, 6.6% against 10.6%, and 6.6% against 8.2%. Over the whole period it lost 28.5% while the basket lost 34%.
That is what a sector sell-off looks like, and in September we warned that this is how equipment makers trade: when the market fears a turn in the chip cycle, their shares "fall first and furthest". What is unusual this time is the order. In 2024 ASML's shares fell because its orders weakened. This autumn they fell while ASML was telling owners that demand exceeded what it could build. The market is pricing a turn that the company does not yet see. One of them will be wrong, and the January guidance for 2027 should tell us which.
Where the price sits. At $1,266.22 the shares are at their lowest close since our X-Ray, 32% below the 2 October high, 18% above the 2025 year-end close of $1,069.86, and 35% above the low of October 2025. In euros, the currency in which ASML earns and reports, the fall is smaller: €1,556 to €1,152, or −25.9%, because the euro weakened 3.4% against the dollar. The market value has fallen from about €592bn to about €438bn ($481bn).
★★ The business ran ahead of the plan — again
A beat, a fourth raised forecast, and a company that says it cannot build enough
On 14 October ASML reported third-quarter net sales of €11.66bn, towards the top of its own range of €11–12bn, and a gross margin of 57.5%, above the top of its range of 55–57%. It sold 121 lithography systems, 17 of them EUV. Its service and upgrade business, which it calls installed base management, brought in €3.29bn, above its own forecast and 68% more than a year earlier. Operating margin was 44.4%. For the fourth time this year the company raised its forecast for 2026, to €44–45bn, about 36% more than in 2025, and it guided to €14.2–15.2bn in the fourth quarter at a gross margin of about 57%.
| Q3 2026 | Value | vs a year earlier |
|---|---|---|
| Net sales (systems / service) | €11.66bn (€8.37bn / €3.29bn) | +55% (€7.52bn) |
| Systems sold (of which EUV) | 121 (17) | 72 a year earlier |
| Gross margin / operating margin | 57.5% / 44.4% | 51.6% / 32.8% |
| Net income / earnings per share | €4.96bn / €12.95 | €2.13bn / €5.49 (incl. a €0.51bn gain, Part IV) |
| Free cash flow (company) | €11.35bn | €0.24bn (incl. ~€5.7bn of down payments, Part IV) |
| Q4 guide: sales / gross margin | €14.2–15.2bn / ~57% | Q4 2025: €9.72bn / 52.2% |
| FY2026 guide | €44–45bn, gross margin ~56% | Jan €34–39bn → Apr €36–40bn → Jul €43–45bn |
The sentence that matters most is not a number. In its release ASML wrote that "our supply as currently planned for both 2027 and 2028 will not fully meet the demand of our customers", that it is "investigating ways to increase capacity for 2028 beyond the 30% increase we previously communicated", and that "some customers are expressing an interest in placing orders with us for 2029 and even 2030". In September we said ASML had a hand on the brake, adding capacity cautiously so as not to be caught by the next downturn with idle factories. It still has; but the brake is now holding back demand, not supply.
What ASML did not say matters too. It gave no forecast of sales for 2027; the chief executive said on the call that there would be no new numbers for 2027 or 2028 (as reported). Its long-standing framework for 2030 is €44–60bn of sales at a gross margin of 56–60%. The 2026 forecast already sits at the bottom of that range, four years early. Either the 2030 framework is too low, which the company will be asked about at its Capital Markets Day on 10 June 2027, or 2026 is a peak, which is what the market fears. The finance chief said price negotiations for new machines were complete, with no effect in 2026, a positive one in 2027 and the full effect from 2028 (as reported).
Memory did the lifting. Memory makers bought 46% to 51% of ASML's systems in each quarter of 2026, against about a third in 2025, and the finance chief expects memory system sales to rise about 75% this year (as reported). Makers of memory chips for AI servers, SK hynix and Samsung above all, have been the fastest-growing customers. That is good news while the AI build-out lasts and a source of risk when it slows: memory is the most cyclical corner of the chip industry, and its spending has fallen hardest in every past downturn.
★ What the third quarter really earned
A one-off gain in the earnings, down payments in the cash, and an order book we can no longer see
Three things in the third quarter look better than they are, and an owner should know them before reading any headline.
The earnings include a one-off gain. Earnings per share of €12.95 include a "gain on equity investments" of €509.7m, about €1.33 a share. ASML's statements show its equity investments rising from €1.32bn to €2.24bn in the quarter: about €406m of new money and about €509m of gain. ASML did not say what the gain relates to. Its largest investment is an 11.1% stake in Mistral AI, the French AI laboratory, bought for €1.3bn in September 2025, and press reports described a new funding round for Mistral in September 2026 in which ASML invested again. Whatever the source, the gain will not recur every quarter. Without it, and assuming it was not taxed, which the quarter's tax rate suggests, earnings per share were about €11.62. That is still 112% more than a year earlier.
The cash includes customers' money. Free cash flow of €11.35bn in one quarter is more than ASML generated in the whole of 2025. About €5.7bn of it is the increase in long-term contract liabilities, which are mostly down payments by customers for machines to be delivered after the next twelve months; the finance chief described them as "underpinning the strong order momentum of our customers" (as reported). Down payments are a sign of demand and a real source of cash, but they are cash ASML owes back in machines. When orders slow, the flow reverses: in the first quarter of 2026, free cash flow was minus €2.6bn. Without the down payments, the quarter's free cash flow was about €5.6bn, still a record.
The order book is no longer visible. Until 2025 ASML disclosed its bookings every quarter, and they were the best early signal of a turn in the cycle. It stopped in 2026. The last figures are bookings of €13.2bn in the fourth quarter of 2025 and a backlog of €38.8bn at the end of 2025, 65% of it EUV. Since then owners have had to rely on words: "order book momentum that continues to increase quarter on quarter this year", according to the chief executive in the release. We believe the words. We would rather have the numbers, and we note that the down payments are, for now, the best number we have.
| Q3 2026 | Reported | Underneath | Read |
|---|---|---|---|
| Earnings per share | €12.95 | ~€11.62 without the €1.33 gain | ◆ Flattered |
| Free cash flow | €11.35bn | ~€5.6bn without ~€5.7bn of new down payments | ◆ Flattered |
| Net cash (Derived) | ~€10.0bn | Cash and short-term investments €12.7bn less ~€2.7bn of bonds; much of it customers' money | ◆ Lent by customers |
| Bookings | Not disclosed | Last: €13.2bn (Q4 2025); backlog €38.8bn | ◆ Blind spot |
| Nine months: earnings per share | €27.67 | Full year on the guide ≈ €42.8 reported, ≈ €41.4 without the gain (Derived) | ✓ In line with consensus |
Why we still read the quarter as strong. Strip out both the gain and the down payments and the quarter is still the best in ASML's history: record sales, a record margin, record service revenue and record cash. The full-year earnings implied by the guidance, about €41 a share without the gain, are in line with what analysts expect (€41.33). And the down payments, which flatter the cash, are themselves the clearest evidence that customers want more machines than ASML can build. The quarter is less spectacular than its headline numbers; it is not less good.
China and the MATCH Act, rechecked
China rose in the quarter; the law did not move; the Dutch government pushed back
China was the risk we spent most time on in September, and the one that behaved least as we expected. In the third quarter Chinese customers took 23% of ASML's system sales, up from 14% in the second quarter and 19% in the first. The finance chief said demand in China "remains very strong" and still expects China to be about 20% of total sales for the year (as reported). In 2025 it was 29.1% of total sales and 33% of system sales.
| China | Share | Basis | Source |
|---|---|---|---|
| 2025 | 33% / 29.1% | System sales / total sales | ASML presentation; 20-F |
| Q1 2026 | 19% | System sales | ASML presentation |
| Q2 2026 | 14% | System sales | ASML presentation |
| Q3 2026 | 23% | System sales | ASML presentation |
| FY2026 expectation | ~20% | Total sales | Finance chief (as reported) |
The MATCH Act. The bill, introduced in Congress on 2 April and advanced by the House Foreign Affairs Committee on 22 April, would give the Netherlands and Japan 150 days to align their controls on older DUV immersion machines, and on servicing them, for five Chinese chipmakers, or face US rules that reach foreign-made equipment (as summarised by IO+). An attempt to attach it to the defence bill failed in July (as reported). We found no vote on the floor of either chamber through 20 November. On 18 November IO+, citing POLITICO, reported that the Dutch foreign ministry had called the bill's reach beyond US borders "unacceptable", and that the prime minister and the foreign-trade minister had raised it in Washington.
Why servicing matters more than sales. A ban on new sales of older machines would hurt; ASML has lived with bans on EUV sales to China since 2019. A ban on servicing machines already installed would hurt more, because service is ASML's steadiest revenue and the Chinese installed base is large. We have seen an estimate of the cost attributed to ASML's finance chief in the press, but we could not find it in any ASML document and do not repeat it. The sensible assumption is that a servicing ban would take a meaningful slice of revenue, and that the Dutch government is resisting it.
Two smaller items. On 16 November the US Commerce Department issued a rule on access to AI computing abroad, which concerns chips, not the machines that make them (as reported). And reports from July said Chinese firms had begun producing their own DUV immersion machines, a handful in 2026, trailing ASML's in performance (TrendForce, citing other reports). Neither is a threat to EUV. Both are reasons why China's share of ASML's sales is unlikely to return to where it was.
The moat, rechecked
Still the only EUV maker; the next machine is still years from mass use
| The claim (September) | Evidence now | Verdict |
|---|---|---|
| Sole maker of EUV machines | 17 EUV systems sold in Q3, at about €240m each (Derived); no rival has a commercial EUV machine. | ✓ Intact |
| Pricing power | Gross margin 57.5% and rising; new prices agreed with customers, effective from 2027 (as reported). | ✓ Stronger |
| Installed base pays for decades | Service and upgrades €3.29bn in Q3, +68% in a year. | ✓ Stronger |
| High-NA is the next monopoly | One High-NA machine sold in Q3; Intel the lead user; Samsung and TSMC mass use around 2028 and 2030 (chief executive, as reported). | ◆ Slow |
| Older machines (DUV) | Canon and Nikon still compete; Chinese makers starting their own (as reported). | ◆ Contested, as before |
We keep the moat at 10. In September we wrote that ASML's EUV monopoly rests on twenty years of work, tens of billions of euros and a chain of suppliers whose skills are nearly as rare as its own. Nothing in eight weeks has changed that, and the evidence of the quarter points the other way from the share price: customers are paying in advance, asking for machines four years out, and accepting higher prices. A moat is measured by what customers do, not by what the market fears.
The one weaker point is High-NA, the next generation of EUV, which prints finer lines. Its adoption is slower than enthusiasts hoped: Intel is using it first, and the chief executive said Samsung and TSMC would use it in mass production around 2028 and 2030 (as reported). That is not a threat to the moat, since no one else makes High-NA either; it is a reason why the next leg of growth may come later than the bull case assumes.
- That earnings grow by about 11% a year for ten years from about €40 a share in 2026; analysts expect 36% in 2027 alone.
- That a pause in AI or memory spending, when it comes, delays ASML's growth rather than reversing it, as in 2023–24.
- That governments do not ban the servicing of ASML's machines already installed in China.
The customers and the cycle
Four customers, 61% of sales — and none of them has cut its spending
| Customer | Latest on spending | Source | Read |
|---|---|---|---|
| TSMC | 2026 capital spending $60–64bn reaffirmed on 15 October; 2026–28 'significantly higher than the past three years'; $40.9bn of spending approved by the board on 10 November. | TSMC releases and board resolutions | ✓ Rising |
| Samsung | Nine-month capital spending up 77%; AI-related spending at a 'historic level' in 2027. | As reported (29 Oct) | ✓ Rising |
| SK hynix | Record third quarter; 2027 spending to be 'significantly' higher than 2026. | As reported (28 Oct) | ✓ Rising |
| Intel | Fourth-quarter forecast below expectations; shares −29% on 30 October. The lead High-NA customer. | As reported | ◆ Weak |
Concentration. In 2025 four customers each bought more than 10% of ASML's sales, together €20.0bn, or 61.2%, up from 53.8% in 2024 (20-F). In the third quarter, customers in Taiwan and South Korea took 63% of system sales. ASML has always depended on a few buyers, because only a few companies can afford leading-edge factories; but the dependence is rising, and it rests on two places, Taiwan and Korea, that carry their own geopolitical risk.
The cycle. Since our X-Ray we have looked for any of ASML's large customers cutting its spending plans, and found none. The fear that sent chip shares down 34% is a fear about the future: that the companies building AI data centres will spend less in 2027 and 2028, and that the chipmakers will then buy fewer machines. The customers' own plans say the opposite. History says that such plans can change within two quarters, and that when they do, equipment orders fall faster than chip sales. Both statements are true. The difference between them is the margin of safety, and the price is now giving us some.
Management & capital allocation
Disciplined operators; a buyback nearly spent at higher prices; an order book no longer shown
The buyback. In January ASML announced a programme to buy back up to €12bn of shares by the end of 2028. By 13 November it had used about €9.4bn of it, 78%, ten months into a programme of thirty-five: about €1.0bn in the first quarter, €1.1bn in the second, €4.1bn in the third and €3.26bn in the seven weeks since our X-Ray. Since 28 September it bought 2.23m shares at an average of €1,462, about 27% above today's euro price; on 5 November, the worst day, it bought 125,000 at €1,258. We do not fault the timing: no one buys only at the bottom. We note the consequence. At the recent pace the programme has five or six weeks left, and the support the company gave its own shares this autumn will end unless the board enlarges the programme, which it has not announced.
Capital allocation is otherwise what we described in September: a dividend that grows every year, buybacks with most of the rest, and investment in capacity before customers demand it. Over 2021–25 ASML returned €31.9bn, 79% of its free cash flow. We lower the management score from 9 to 8 for one reason: the decision to stop disclosing bookings. It removed the best early-warning number an owner had, in a business whose shares trade on the cycle, and at a time when the cycle is the question. We would welcome it back.
Insiders. ASML is a Dutch company and its managers report their dealings to the Dutch regulator, not on the US forms we use for American companies. We have nothing from those registers to report here and draw no conclusion from that.
The numbers
In euros; the cash, the margin, and six faults in our feed
Nine months. In the first nine months of 2026 ASML sold €29.76bn, at a gross margin of 55.0%, and earned €10.64bn, or €27.67 a share. On the company's guidance the full year will bring €44–45bn of sales and earnings of roughly €43 a share, €41 without the third-quarter gain. Trailing twelve-month earnings are €35.04 a share, €33.71 without the gain.
The cash. Free cash flow over twelve months was about €21.0bn on ASML's definition, about 4.8% of the market value; without the €5.7bn of new down payments, about 3.5%. Over the five years to 2025 ASML turned 110% of its net income into free cash flow (€40.4bn of free cash against €36.5bn of profit), which is why we treat its earnings as close to owner earnings. Its capital spending, €1.6bn in 2025, is small for a company of its size, because its suppliers and its customers carry much of the heavy investment.
The balance sheet. At 27 September ASML held €12.7bn of cash and short-term investments against roughly €2.7bn of bonds: about €10.0bn of net cash, or €26 a share. As Part IV explains, much of it is customers' money paid in advance. Interest costs are trivial: €118m in 2025. There is no financial risk here; there is only the reminder that cash which arrives as down payments can leave as quickly when orders slow.
| What the screen shows | What is wrong, and the number to use |
|---|---|
| Dividends: 2.138 in July, 2.096 in October — a cut | Amounts converted to dollars without a currency label. ASML declares in euros: <b>€1.88 both times</b>, unchanged. Growth and streak tests on these values measure the exchange rate. |
| Quarterly results: Q3 2025 listed twice | One quarter, two rows (28 and 30 September 2025). We flagged it in September; still unfixed. |
| Quarterly debt and net debt | Leaves out short-term debt and ignores €4.7bn of short-term investments. This is the source of our own '€10bn' error in September. Use the annual record or the filing. |
| Interest expense: 0; interest cover: 0 | The 20-F shows <b>€118.3m</b> for 2025. Interest cover is in the hundreds, not zero. |
| Sales by product and region, 2025: €38.5bn and €38.0bn | Junk and duplicate rows inflate the totals by 16–18%; 2024 is labelled in dollars though the values are in euros. Total 2025 sales were <b>€32.67bn</b>; China was 29.1%. |
| Analysts' mean target: $2,421.83 | Three targets, no analyst count, stale. StockAnalysis shows <b>~$2,105</b> from 41 analysts, with a low of $848. |
The dividend
Small in yield, fast in growth, and covered several times over — tested
ASML pays a dividend every quarter: three interim payments and a final one approved at the annual meeting. For 2025 it paid €7.50 a share (three interims of €1.60 and a final of €2.70), 17% more than for 2024. For 2026 the interims rose to €1.88, 17.5% more; the second was paid on 4 November. At four times €1.88 the yield is about 0.65% at today's price. US holders receive it after Dutch dividend tax. It is a small dividend, but one that has grown about 23% a year since 2017.
| Test | ASML | Result |
|---|---|---|
| 1. Covered by earnings | 30% of 2025 earnings; about 18% of 2026's expected earnings at the €1.88 rate | ✓ Pass |
| 2. Covered by free cash flow | 23% of 2025 free cash flow; with buybacks, 77% | ✓ Pass |
| 3. Growing | €1.40 for 2017 → €7.50 for 2025 (≈23% a year); 2026 interims +17.5% | ✓ Pass |
| 4. Never cut | Eight consecutive increases since 2017 (earlier years not checked) | ✓ Pass |
| 5. Balance sheet behind it | Net cash ~€10bn ≈ 3.5 years of dividends, much of it down payments | ✓ Pass |
| 6. A stated policy | 'A sustainable dividend per share that will grow over time, paid quarterly' (20-F) | ✓ Pass |
The one point to watch is not the dividend but the total returned. In 2023 ASML returned slightly more than its free cash flow, and in the first nine months of 2026 it returned 85%. In a downturn the buyback is the part that would shrink, and the dividend, at under a quarter of free cash flow, would not need to. For income investors ASML is a dividend-growth holding, not a yield; for everyone else the dividend is a sign of discipline rather than a reason to buy. In our feed the dividend appears to have been cut in October. It was not; see fault one in Part IX.
★ Valuation
About 20 times next year's earnings for the only machine of its kind
Our method. In September we set our range as 25–30 times the earnings expected for 2027, and our "both hands" level at 22.5 times. Refreshed for today's estimates and exchange rate, that method gives a range of about $1,545–1,853 and a both-hands level of about $1,390. This time we also value ASML the way we value most businesses: ten years of owner earnings, which for ASML are close to reported earnings, discounted at 10% a year, with a value for the years after, plus net cash. The three cases are three paths for the chip cycle.
| Path | 2027 owner earnings | Growth after | Terminal | Value today |
|---|---|---|---|---|
| AI and memory pause | €45 a share | 5% a year | 3% | ~$830 |
| Central case ◆ | €56 (consensus) | 12% to 2031, 7% to 2036 | 3.5% | ~$1,430 |
| The wave runs on | €60 | 15% to 2031, 10% to 2036 | 4% | ~$1,955 |
| What $1,266 implies | From €40 (2026, without the gain) | ~11% a year for ten years | 3.5% | $1,266 |
What is priced in? At $1,266 the shares assume that ASML's earnings grow about 11% a year for ten years from this year's level, without the one-off gain. At our September price the same arithmetic implied about 15%. Analysts expect 36% growth in 2027 alone, and the company says it cannot meet demand in 2027 or 2028. The price therefore assumes either a sharp slowdown after 2028 or a down year before then. Both are possible in this industry; neither is in the company's guidance or in its customers' plans.
Why not wait for lower? Because the business is worth more than the price in two of our three cases, and the central case is fairly robust. If growth after 2027 is 10% rather than 12%, the central value falls to about $1,340; if 2027 earnings are €50 rather than €56, to about $1,280; if the terminal growth is 3% rather than 3.5%, to about $1,365. All are above or close to today's price. Only the pause case, in which 2027 earnings fall back to about this year's level and growth stays at 5%, is far below it, and that case would require ASML's customers to abandon the plans they announced this autumn.
Our level. In September we said to keep powder dry and to buy with both hands near $1,365. That price came, and it came for reasons outside ASML. So we keep the promise: buy now. But equipment shares have fallen much further than this in past downturns, and the evidence for a turn would arrive in January, not before. We would therefore keep the last part of a position for a lower price: about $1,150, 20% below our central value, 18.6 times 2027 expected earnings and a price that implies about 10% growth a year. Our level moves from $1,365 to $1,150 not because the business is worth less but because the first level has been used.
Risks, regulation & controversies
Verified 22 November 2026 against ASML's 6-Ks, the 20-F and attributed reports
Legal and regulatory. We found no material lawsuit against ASML. The regulatory risks are the export rules of the Netherlands, Japan and the United States, and the MATCH Act, which had not been voted on by either chamber of Congress by 20 November. ASML's own release lists, among its risks, "restrictions on shipments of systems, including ordered systems, under export controls".
What we did not use. Press reports have attributed to ASML's finance chief an estimate of what a servicing ban would cost, and have described a Chinese EUV light-source prototype. We could not confirm either in a primary source, and neither appears in this report's numbers.
| Test | Today | What would change our mind |
|---|---|---|
| Q4 results and the first word on 2027, late January | FY2026 €44–45bn | Talk of 2027 sales flat or lower would mean the market saw the turn first; we would stop at the first stage. |
| Down payments (long-term contract liabilities) | €9.3bn, +€5.7bn in Q3 | A fall would be the first sign of orders slipping, since bookings are not disclosed. |
| Customers' 2027 spending plans | TSMC, Samsung, SK hynix: higher | A cut by any of the three would test our central case. |
| MATCH Act | Not voted | A servicing ban enacted and matched by The Hague would cut our values by roughly the Chinese service revenue. |
| The buyback | ~€2.6bn of €12bn left | An enlarged programme at today's prices would show the board agrees with us. |
| The price | $1,266 | Below $1,150 with the tests intact: complete the position. |
In September I told you that ASML was the purest monopoly on our board, that you should accumulate it slowly, and that you should keep some powder dry for the next cyclical scare. If the scare offered the shares near thirteen hundred and sixty-five dollars, I said, buy with both hands. Eight weeks later they are at twelve hundred and sixty-six. A reader who bought in September is more than a quarter behind, and deserves to know why.
The scare came, but not the cycle. In those eight weeks ASML sold more than it had forecast, at a higher margin, raised its forecast for the year for the fourth time, and told owners that it cannot build enough machines for its customers in 2027 or in 2028. Its customers paid it more than five billion euros in advance. None of them cut its spending. The shares fell because every chip share fell, by a third, on the fear that the companies building AI data centres will soon spend less. ASML fell less than the rest.
I also owe you five corrections, and I have made them in this report. The most important: China was not a third of ASML's sales last year but twenty-nine per cent, and the net cash I gave you came from a feed that left out part of the debt. Neither changes the verdict. Both are the kind of error a careful owner should not have to find for themselves.
I will not pretend the quarter was quite as good as its headline. Part of the profit was a one-off gain on an investment; part of the cash was customers' money for machines not yet built; the company no longer tells owners how many orders it has taken; and the buyback that supported the shares this autumn is nearly spent. A wise owner reads those things and is not alarmed by them, but does not forget them either.
So I keep the promise. The price I named has come, and it has come for reasons that have nothing to do with the business. Buy now. At about twenty times next year's expected earnings you are paying an ordinary price for an extraordinary business. But keep the last part of the position for eleven hundred and fifty dollars or lower. If the turn the market fears is real, January will show it, and the shares will offer a lower price. If it is not, you will own most of what you wanted at a price you will be glad of.
— The Buffett Lens. When the market sells a monopoly with its neighbours, buy the monopoly; keep a little back for the day the neighbours turn out to be right.
Add Again Below $1,150The purest monopoly on our board, sold with its sector while its business ran ahead of plan. ★ At $1,266 the price is below our September both-hands level ($1,365) and our central value of ~$1,430, within a range of ~$830 (AI and memory pause) to ~$1,955 (the wave runs on). Buy now; keep the last part for below $1,150, about 20% under our central value. Q4 results and the first word on 2027 in late January.
Articles that take this apart
MoatsA shortage is not a moatScarcity and pricing power produce identical numbers for a year or two. Then capacity arrives, and only one of them is still there.
MarketWho actually gets paid in the AI build-outEnormous sums are moving. Following where they settle, rather than where they are announced, separates the businesses collecting cash from the ones committing it.
PortfolioThe risk you cannot diversify awaySome risks get smaller when you own more things. A few do not get smaller at all, and for those there is exactly one tool — and it is not diversification.