Semiconductor capital equipment. ASML / KLAC. Run 14 September 2026. Built from ASML 20-Fs FY2023 to FY2025 (last filed 25 Feb 2026) and 6-Ks through 15 Jul 2026; KLA 10-Ks FY2024 to FY2026 (last filed 6 Aug 2026), Investor Day 12 Mar 2026, Q4 FY2026 call 28 Jul 2026 and Citi TMT 9 Sep 2026; ASML Investor Day Nov 2024; company press releases. Events swept through 14 September 2026; most recent event checked: the ASML and TSMC 12-inch photomask initiative for High NA EUV, announced 8 September 2026. No FX conversion is applied: every comparison is a ratio, a growth rate or a share of total, and absolutes stay in each company's reporting currency (ASML EUR, KLA USD). Not a valuation and not a recommendation.
ASML's forward engine is revenue per build slot, and on 15 July 2026 it put a slot count in public for the first time: it plans to add 30% to a 2026 low-NA EUV capacity "of around 65" for 2027, the same 30% to a DUV immersion capacity "of around 130," and is investigating a further 30% for 2028. The mechanism is not volume. ASML shipped 327 lithography systems in FY2025 against 418 in FY2024 and still grew net system sales 12.4%, because it swapped mature machines for EUV ones and repriced the NXE line 26.9% inside a single generation (inferred from the FY2025 20-F unit table; ASML publishes no ASP). That engine has already outrun the company's own long-term model: FY2026 guidance of EUR 43 to 45bn sits at or above the low end of the November 2024 scenario range of EUR 44 to 60bn for 2030, four years early, which is why that model is treated here as stale rather than as evidence. KLA's engine is broader and slower, resting on wafer equipment reaching USD 215bn by 2030, process-control intensity rising from 7.4% to about 9% of it, and share within process control rising from 56.5%. Two of those three legs are outgrowth of the market, and ASML has only the first. The call goes to ASML on rate and concentration of force, not on durability, and its growth is the more violent of the two: one cell at 35.5% of sales grew 39.4% in FY2025 while another at 5.3% of sales fell 44.6% in the same year.
The growth winner and the margin winner are different companies, and the reason they differ is the single most important structural fact in this duel. KLA converts at 60.7% gross margin and 40.0% EBIT margin on a three-year average against ASML's 51.8% and 33.1%, with no crossover in any of the three years, and the 6.9-point EBIT gap reconciles exactly: 8.9 points of gross margin, plus 2.8 points of lower R&D intensity, less 4.9 points of heavier SG&A. The conversion mechanism is vertical position, not price. ASML holds the better price lever of the two, having repriced NXE 26.9% in a year while KLA's CFO said on 28 July 2026 that "it's pretty hard to go back to your customers after you've taken orders and start to change prices on those orders," and ASML still loses the margin comparison by nine points, because it buys the physics it sells. EUR 4,406.9m of its FY2025 cost of sales, 28.6%, went to Carl Zeiss SMT alone (inferred: 4,406.9 divided by 15,409.3, both disclosed in the FY2025 20-F), under a contract priced by "a variable pricing model determined by the annual financial performance of both ASML and Carl Zeiss SMT GmbH." KLA makes its own light sources, its own 300kg catadioptric objectives, its own sensors and its own image computers. In plain terms: ASML is a monopolist that has to share, and KLA is an oligopolist that does not. One is a compounder whose engine is leased; the other is a fortress that still has ground to take.
ASML's worst exposure is that its supply risk and its credit risk are the same counterparty. Carl Zeiss SMT is the sole and exclusive supplier of every lens, mirror, illuminator and collector, produced at two German sites, and ASML states that if Zeiss stopped it "would effectively cease to be able to conduct our business." ASML simultaneously carries EUR 3.92bn of exposure to it, being EUR 1,907.5m of loans receivable, EUR 1,191.9m of advance payments and a EUR 822.6m equity stake, with the 2025 restated framework making those loans interest-free and repayable out of Zeiss's revenue, plus operating losses that "cannot be quantified." The 2027 capacity addition runs through the same factories. KLA's worst exposure is different in kind: it is a percentage of somebody else's capex with no lever on the denominator, and if wafer equipment stalls its share-gain leg is worth roughly USD 3bn of the USD 13bn increment in its 2030 model, real but not enough to carry it alone. In a demand trough ASML breaks first, because its revenue is price-per-slot levered rather than volume-levered, roughly 47% of its cost of sales is bought in on a contract that shares upside and does not fall proportionally, and it negotiates with four counterparties who know they are 61.2% of its revenue. KLA enters the same trough with 23% of revenue in services that are more than 80% contract-based on contracts longer than three years, tools that recur in revenue 95% of the time over any two-year window, and a business that has had "only one down year, I think, in 25 years" (Higgins, 9 Sep 2026). KLA breaks first only in a competitive shock, and competitive shocks arrive slowly.
These six cells carry essentially all of the combined economics: three cells account for 92.2% of ASML's revenue and two for 91.8% of KLA's. Every score is argued in the tabs below.
| Cell (product x region) | ASML | KLA | Why (one clause, sourced) |
|---|---|---|---|
| Leading-edge patterning systems, Taiwan and Korea | 5 | 0 | ASML is the only supplier of EUV and names no EUV competitor at all (FY2025 20-F); KLA makes no lithography |
| Leading-edge process control, Taiwan and Korea | 2 | 5 | KLA 56.5% share and "6.5x our nearest competitor" (Khan, 12 Mar 2026) against ASML's EUR 824.6m of metrology and inspection, 2.5% of sales |
| Mainstream and mature systems, China | 4 | 3 | ASML holds immersion but its mature lines fell 44.6% in FY2025; KLA's China revenue was flat at USD 4.05bn and it is blocked from fabs competitors can ship into (Higgins, 28 Jul 2026) |
| Advanced packaging process control, global | 2 | 5 | KLA number one from "number four in a two-player market," advanced packaging process control systems revenue of roughly USD 1.1bn in calendar 2026, up more than 70% (Wallace, 28 Jul 2026); ASML participates only through mature packaging lithography |
| Memory equipment, Korea | 4 | 4 | ASML Korea 25.0% of sales with memory systems of EUR 8,420.2m; KLA Korea 13.5% but DRAM revenue rose 3x while DRAM wafer equipment rose 2x (Khan, 12 Mar 2026) |
| Installed-base service, global | 4 | 5 | ASML grew 26.2% but publishes no service-versus-upgrade split and shifted system revenue into the line in FY2025; KLA is more than 80% contract-based with 95% two-year recurrence and 17 consecutive years of growth |
Scores are anchored to the exhibits in the three tabs; a score with no exhibit behind it does not ship. Scores are per cell and are not summed - the three lenses get their verdicts in the Three Answers above, not here.
Before any table, three normalisations. ASML's fiscal year is the calendar year; KLA's ends 30 June, so KLA's most recent reported year sits six months later in the same upcycle and its reported growth is flattered by that. No FX conversion is applied anywhere: every comparison below is a ratio, a growth rate or a share of each company's own total, and absolutes stay in reporting currency. And the product labels are mapped, because they do not match: ASML's EUV, ArF immersion, mature DUV, metrology and inspection, and installed-base management map onto KLA's wafer inspection, patterning, specialty semiconductor process, PCB and component inspection, and services.
The cross-tabulation an 80/20 cell analysis wants does not exist in either company's disclosure. ASML reports one reportable segment and discloses systems by technology, by end use and by geography, but never crossed. KLA reports three segments and discloses product category and geography, never crossed. Every product-and-geography cell here is therefore marked (inferred) with its logic stated, and that non-disclosure is itself a finding rather than a research failure.
What the cells show is an almost total absence of overlap. ASML's two largest product cells, EUV at 35.5% of sales and mainstream immersion at 31.6%, have no KLA counterpart at all, because KLA makes no lithography. KLA's largest cell, process control at 68.8% of its revenue, faces an ASML line that is 2.5% of ASML's sales. The two companies contest one cell and are eleven times apart inside it.
| Cell | ASML rev (% of total) | growth | KLA rev (% of total) | growth | Margin signal and leader |
|---|---|---|---|---|---|
| Leading-edge patterning (EUV) | EUR 11,602.7m (35.5%) | +39.4% | ND | ND | EXE explicitly dilutive to gross margin (ASML FY2025 20-F). ASML, uncontested |
| Mainstream immersion litho | EUR 10,311.4m (31.6%) | +6.7% | ND | ND | ND, no segment margin disclosed. ASML, uncontested |
| Process control systems | EUR 824.6m (2.5%) | +27.7% | USD 9,337.6m (68.8%) | +11.2% | KLA Semiconductor Process Control segment profit 44.8% of segment revenue (FY2026 10-K). KLA, decisively |
| Mature, specialty and adjacent | EUR 1,735.6m (5.3%) | -44.6% | USD 1,115.9m (8.2%) | +3.5% | KLA's PCB segment swung to USD 108.0m profit in FY2026 from a USD 281.2m loss. KLA on trend |
| Installed-base service | EUR 8,193.0m (25.1%) | +26.2% | USD 3,125.9m (23.0%) | +16.5% | ASML ~50.8% (inferred: service revenue less cost of service, both disclosed); KLA lower than corporate but accretive to operating margin (Higgins). ASML on rate, KLA on quality |
| Region | ASML, % of FY2025 net sales | KLA, % of FY2026 revenues |
|---|---|---|
| China | 29.1% (EUR 9,519.7m) | 29.8% (USD 4,048.4m) |
| Taiwan | 25.5% (EUR 8,337.9m) | 26.8% (USD 3,643.7m) |
| Korea | 25.0% (EUR 8,159.6m) | 13.5% (USD 1,833.8m) |
| North America | 12.5% (EUR 4,089.1m) | 13.0% (USD 1,757.3m) |
| Japan | 4.3% (EUR 1,420.9m) | 6.7% (USD 915.1m) |
| Europe including Israel | 1.6% (EUR 529.0m) | 5.4% (USD 726.7m) |
| Rest of Asia including Singapore | 1.9% (EUR 611.1m) | 4.8% (USD 654.4m) |
Three crossings can be built honestly. First, ASML's EUV business does not exist in China: EUV has never been licensed for export there, so ASML's EUR 11,602.7m of EUV revenue sits entirely inside its EUR 23,147.6m of non-China sales, making EUV 50.1% of ASML's non-China revenue (inferred: EUV revenue divided by total net sales less China net sales, all three disclosed). ASML's China business is DUV, mature nodes and service. Second, the mature-line collapse is a China normalisation rather than a licence event: the mature lines fell 44.6% while China sales fell only 6.6%, and ASML states its 2025 China DUV business "turned out to be stronger than anticipated." Third, KLA's China cell has a hard ceiling with a blocked component: its China revenue was flat in FY2026 at USD 4,048.4m against USD 4,042.6m while total revenue grew 12%, and Bren Higgins said on 28 July 2026 that "a number of our competitors have been able to ship into fabs in China that we haven't been able to ship into."
ASML, EUV. The driver is revenue per build slot. ASML sold 327 lithography systems in FY2025 against 418 in FY2024 (20-F unit table: 535 total less 208 metrology and inspection, against 583 less 165) and still grew net system sales 12.4%, with implied NXE pricing moving from EUR 187.1m to EUR 237.4m per unit, up 26.9% (inferred; ASML publishes no ASP). The forward driver is the capacity number it gave on 15 July 2026: plus 30% on roughly 65 low-NA EUV slots for 2027, plus 30% on roughly 130 DUV immersion slots, with another 30% under investigation for 2028. The cost of growing there is that the line runs through a supplier: "The number of lithography systems we are able to produce is limited by the production capacity of one of our key suppliers, Carl Zeiss SMT." Growth in this cell is entirely organic.
KLA, process control. Three legs, each quantified by management at the 12 March 2026 Investor Day. Wafer equipment to USD 215bn plus or minus 20bn by 2030, about a 12% compound rate. Process-control intensity from 7.4% of that to about 9%, driven by larger die ("if you have 600 chips on a wafer and a 1% yield loss, and you do nothing else but increase the chip size, and you have 60 chips per wafer, that equivalent process will have a 10% yield loss"), by more designs per node, by buried defects in gate-all-around structures, and by high-bandwidth memory, where "we're seeing intensity levels that are as high as high-end logic." And share within process control, 56.5% and "6.5x our nearest competitor," up two points since 2021. The second and third legs are outgrowth of the market; ASML has the first and a version of the second, and cannot have the third.
KLA, advanced packaging. From 10% share in 2020 and "number four in a two-player market" to number one, with calendar 2026 revenue of roughly USD 1.1bn, up more than 70% and "almost two times faster than the advanced packaging market." The driver is specification tightening: hybrid-bonding overlay requirements are "10x tighter" than micro-bump, which drags front-end tools into a market that had been served by inspection a generation behind. The cost of growing there is margin. "Packaging generally carries a lower gross margin on great gross margin dollar opportunity" (Higgins, 9 Sep 2026), and only "maybe 20% or so" of that USD 1.1bn is the higher-end front-end tooling today.
ASML, installed-base management. EUR 8,193.0m and up 26.2%, now a quarter of sales. But ASML states that in 2025 it "carried out a significant number of TWINSCAN NXE:3800E field upgrades, which resulted in a substantial portion of EUV system revenue being shifted to installed base revenue," and it does not split service from field upgrades anywhere in the filing. The line is one number with no composition.
Both sell direct. Neither uses distributors, wholesalers or tenders, so the usual channel question is moot and the real question is repricing cadence. ASML prices value-based under volume purchase agreements of up to five years that its own revenue policy says contain "discounts, free goods or services and credits," on 15 to 45 day terms with cancellation penalties, and it has repriced inside a generation. KLA reprices only at new-product introduction: "When we bring out new products, that's when new pricing decisions get made. With existing products, it's a little bit more around other ways where we can share in some of the increased value" (Wallace, 28 Jul 2026). Pressed on the same call about why KLA could not do what ASML is doing on EUV pricing, Higgins was blunt: "It's pretty hard to go back to your customers after you've taken orders and start to change prices on those orders."
The consequence shows up in the accounts. KLA is absorbing a memory input-cost increase that Higgins put at around 100 basis points of gross margin on 28 July 2026 and at "slightly ahead of 100 basis points" on 9 September 2026, which it cannot pass through on booked orders and expects to persist through 2027. ASML's equivalent cost inflation is shared with its largest supplier by contract, because the Zeiss purchase price runs on a variable model set by the annual financial performance of both parties. KLA's offset is cadence rather than contract: its cycles are short and high-mix, so the repricing window comes around often.
| ASML | KLA | |
|---|---|---|
| Critical single source | Carl Zeiss SMT, sole and exclusive supplier of all lenses, mirrors, illuminators and collectors, from two sites, Oberkochen and Wetzlar | Optical components generally; calcium fluoride lead times "18 to 24 months or more" (Higgins, 9 Sep 2026) |
| Size of that dependency | EUR 4,406.9m of purchases in FY2025 = 28.6% of cost of sales (inferred: 4,406.9 / 15,409.3, both disclosed) | ND - KLA discloses no supplier concentration |
| Vertical integration | Light source in-house (Cymer, 1,000W demonstrated April 2025); optics bought | Light sources, 300kg catadioptric objectives, custom sensors and image computers all in-house |
| Stated consequence of failure | "we would effectively cease to be able to conduct our business" (FY2025 20-F) | No equivalent statement; risk framed as lead time |
| Mitigation | Funds the supplier's R&D (EUR 22.5m for High NA in 2025) and its capex above thresholds; ~5,100 suppliers, 272 business-critical = 89% of product spend | About USD 250m invested in supplier capacity; eight quarters of visibility; "we don't cancel orders" (Higgins, 12 Mar 2026) |
In lithography, ASML names only Canon and Nikon and only "in respect of DUV systems." For EUV it names no competitor at all, framing the threat instead as new entrants "driven by the ambition of self-sufficiency" and as substitution by "alternative technological solutions." ASML is a share-taker in EUV by default and a share-donor at the mature end it is exiting by choice. In process control, KLA claims 56.5% share and 6.5 times its nearest competitor, implying the number two is below 10%, and it names no process-control competitor anywhere in its 2026 disclosures. Applied Materials appears only in an analyst's question at Citi on 9 September 2026, to which Higgins replied: "If I had a nickel for every time one of our competitors talked about gaining share in the industry, I'd have a lot of nickels." Named-competitor share data: ND. ASML, for its part, does name KLA: it competes with "Applied Materials Inc. and KLA-Tencor Corporation" in applications that support or enhance complex patterning solutions and compete with ASML's own offerings, which ASML says are "a significant part of our business" - the only place in either company's filings where the two appear in one sentence.
The event that changes the map is six days old. On 8 September 2026 ASML and TSMC announced an industry initiative to move High NA EUV from 6-inch to 12-inch photomasks, targeting a pilot line in 2031 and full system readiness in 2033, with TSMC starting High NA high-volume manufacturing in 2030. A mask-format change reshapes the mask infrastructure, and mask shops are the single highest process-control intensity segment KLA serves: "almost half of a spend for mask shop is process control" (Wallace, 12 Mar 2026). It is simultaneously KLA's largest long-dated opportunity and its largest long-dated obsolescence risk, and neither company has sized it.
For ASML's EUV cell in Taiwan and Korea, the binding risk is that ASML cannot build enough, not that a rival builds instead, and the 2027 plus-30% runs through Zeiss's factories. For ASML's mature DUV cell in China, already down 44.6% in a year, 10 November 2026 is a double date: the US BIS Affiliates Rule is reimposed after its one-year suspension and China's rare-earth control suspension expires, within days of each other. ASML says the Affiliates Rule "currently has no impact on our business operations" precisely because it is suspended. Across all of ASML, one reportable segment and no service-versus-upgrade split mean an investor cannot see where a downturn lands. For KLA's process-control cell in Taiwan and Korea, one customer was about 19% of FY2026 revenues, unchanged, and 2026 intensity ran below model because KLA could not ship: "if we could have built more, we could have shipped more." For KLA's advanced-packaging cell, the fastest cell is the most dilutive. For KLA's service cell in China, export controls bite the most resilient leg: service growth landed at the bottom of its range because it "had some disruption related to access to fabs" (Higgins, 12 Mar 2026).
EUV monopoly. Durability High. ASML names no EUV competitor in its FY2025 20-F; the competition risk factor cites only Canon and Nikon, and only for DUV. The barrier is not capital but two decades of a light source and an optics supply with one source on earth. A well-funded entrant needs years, not money.
Installed-base upgrade path. Durability Medium. ASML converted NXE:3800E field upgrades into EUR 8,193.0m of installed-base revenue, up 26.2%. But it publishes no split between contracted service and discretionary upgrade, so the durable portion cannot be verified. Rated Medium for that reason alone.
Regulatory asset by accident. Durability Medium. Export licensing has frozen the Chinese leading-edge market shut, protecting ASML from the only entrant that is actually funded. This is a moat ASML does not own and a government can remove in either direction.
Share plus an R&D gap. Durability High. 56.5% share and "6.5x our nearest competitor," implying number two below 10% (Khan, 12 Mar 2026); "typically, we're investing more than their revenue, just in R&D" (Wallace, same day), a spread that "was 4x seven years ago." Money alone does not close a gap where the leader's R&D exceeds the follower's revenue.
High-mix application knowledge. Durability High. "Process control is tricky to get into because it's a very high-mix, low-volume market. Unlike litho, frankly, is much higher kind of similar tool, higher volume" (Wallace, 28 Jul 2026). One product division alone carries more than 30 models across five configurations, 1,200 unique assemblies and 25,000 specifications, supported by 1,600 to 1,700 applications engineers.
Service installed base. Durability High. More than 57,000 tools at more than 4,000 facilities; more than 80% of service revenue on contracts longer than three years; "every one of those individual tools over a two-year period recurs in revenue 95% of the time" (Lorig, 12 Mar 2026); median time to tool retirement has gone from about four years in 2000 to more than 24 years in 2025, and the 1995 product class still produces more than USD 100m of annual recurring service revenue.
ASML: four customers each exceeded 10% of net sales in FY2025, together EUR 20.0bn or 61.2% of revenue, up from 53.8% in FY2024 and 53.9% in FY2023; the largest alone was EUR 7,796.7m or 23.9%, and the three largest were 35.4% of receivables. Access is not contested because there is no alternative supplier, so price is set by value, under five-year volume purchase agreements that ASML's own revenue policy says contain discounts, free goods and credits. Pricing power has been exercised: implied NXE pricing rose 26.9% in one year (inferred).
KLA: one customer at approximately 19% of FY2026 revenues, unchanged from FY2025 and up from 13% in FY2024; "our top five customers drive a lot of our business" (Higgins, 9 Sep 2026). Access here is contested, and KLA's stated defence is a refusal to buy share: "We constrain market share by gross margin. You can have high market share, but if you're buying the market, that's not great as an investment" (Wallace, 12 Mar 2026). Pricing power is exercised only at new-product introduction.
ASML: single-source, exclusive and existential. Carl Zeiss SMT is "our sole supplier of lenses, mirrors, illuminators, collectors and other critical optical components," ASML is Zeiss's single customer for optical columns, production sits at two German sites, and if Zeiss stopped ASML "would effectively cease to be able to conduct our business." ASML's balance-sheet exposure to that same counterparty at 31 December 2025 was EUR 1,907.5m of loans receivable, EUR 1,191.9m of advance payments and a EUR 822.6m equity stake, EUR 3.92bn in all, plus "the risk of any future operating losses which cannot be quantified." The 2025 restated framework made the loans interest-free with repayment tied to Zeiss's revenue. Pass-through: partial, by contract - the purchase price runs on a variable model determined by the annual financial performance of both parties, which shares cost shocks and shares upside away.
KLA: deliberately sole-sourced at the top tier, but with no single point of total failure. "For high-value components, we don't compete against our competitors because the relationships are mostly sole sourced. When industry strained, I'm not gonna have to worry about an allocation necessarily" (Higgins, 12 Mar 2026). The hardest components are made in-house: the DUV broadband light source, the 300kg catadioptric objectives ("There's one place in the world that can make those things"), custom sensors and custom image computers. Dual-sourcing coverage: Medium by design. Pass-through: weak on booked orders, which is exactly what the roughly 100 basis point memory-input cost absorption shows.
| Cell | Route control | Pocket price | Continuity | Outcome, share and margin | Confirming KPI |
|---|---|---|---|---|---|
| Leading-edge patterning, Taiwan and Korea | ASML direct, sole supplier. KLA no presence (ND) | ASML above: reprices within a generation, NXE up 26.9% (inferred). KLA not applicable | ASML capped by Zeiss capacity; plus 30% planned for 2027 | ASML share flat at roughly 100%, nothing left to gain; margin up as EUV mix rises and EXE dilution fades | ASML 2027 low-NA EUV units of 85 or more |
| Process control, Taiwan and Korea | ASML marginal, EUR 824.6m. KLA direct, 56.5% share | ASML below: subscale, selling against a 6.5x leader. KLA above: "we constrain market share by gross margin" | ASML not constrained (small). KLA constrained: 12-month lead times, 18 to 24 on Gen4 and Gen5 | KLA share up, plus two points since 2021; margin up unless Applied Materials forces a response | KLA process-control share of 58% or more at its next update |
| Installed-base service, global | Both direct. ASML composition undisclosed; KLA more than 80% contracted | ASML ND, roughly 50.8% gross margin (inferred). KLA below corporate but accretive to operating margin | Both export-restricted in China; KLA reports "disruption related to access to fabs" | ASML revenue up faster, plus 26.2% against plus 16.5%, but quality unverifiable; KLA slower and provable | ASML publishing a service-versus-upgrade split |
Where the two lenses disagree. ASML grows fastest exactly where its power is most complete, so on the first cell the growth map and the power map agree. They part company on the third. ASML's installed-base line is its second-largest and second-fastest, and it is the one line where ASML's disclosure is weakest: a single combined number, no split, in a year ASML itself says shifted system revenue into it. KLA's equivalent line is smaller and slower and is the best-documented asset anywhere in this duel. Fast but unverifiable against slower and provable is the most valuable finding of this run, because it is the difference between a thesis you can monitor and one you can only hold.
1. Vertical position. Impact: Major. ASML buys its optics from a monopolist that prices off ASML's own profit; KLA makes its optics and light sources in-house. This is the entire 8.9-point gross-margin gap. To close it ASML would have to internalise Zeiss or rewrite the variable-pricing model; it has owned 24.9% of Carl Zeiss SMT Holding since June 2017 and the model still stands. Realistic timeline: a decade, or never.
2. Room left to gain. Impact: Major. ASML has roughly all of EUV and nothing left to take; KLA has 56.5% of a market it leads by 6.5 times, with 43.5 points still in play. This is the only structural growth leg either company holds that does not depend on the customers' capex budget. It is reversible by a funded competitor, which is flip condition two.
3. Repricing cadence. Impact: Moderate, and it runs in ASML's favour. ASML reprices within a generation; KLA only at new-product introduction and explicitly not on booked orders. To change it KLA would only need different terms at order intake, which is a policy rather than a moat and could move in one product cycle of about two years if customers tolerated it. Because it is closable with a decision rather than a decade, it is not weighted as a durable advantage.
Both companies present costs by function, so no by-nature reconstruction is needed and all five ratios are defensible. Two mapping notes, applied to every year. ASML reports under US GAAP in EUR with cost of system sales and cost of service and field option sales summing to cost of sales, and it reports an explicit income-from-operations line; no restructuring or one-off is itemised across FY2023 to FY2025, so nothing is excluded. KLA reports under US GAAP in USD but presents no operating-income line at all, running from revenues straight to income before taxes after interest and other income, so EBIT here is revenues less cost of revenues, R&D and SG&A. KLA recorded goodwill and intangible impairments of USD 289.5m in FY2024 and USD 239.1m in FY2025 in its PCB segment, and nil in FY2026; because the company itemises them they are shown both ways, and the ex-impairment line is the one comparable with ASML, which had no such charge.
Fiscal years are offset by six months, more than a quarter, so it is stated rather than buried: KLA's window sits six months later in the same upcycle, which flatters its growth and, to a lesser degree, its margins. Ratios are less sensitive to this than absolutes, and no ratio gap below is small enough for the offset to explain it.
| % of sales, 3-year average | ASML | KLA | Gap (KLA less ASML) | What drives it |
|---|---|---|---|---|
| COGS | 48.2 | 39.3 | -8.9 | ASML buys in its hardest components, one supplier being 28.6% of FY2025 cost of sales; KLA makes light sources, objectives, sensors and image computers itself |
| R&D | 14.7 | 11.8 | -2.8 | ASML funds a single-generation platform transition in High NA plus its supplier's R&D; KLA spreads R&D across more than 30 models on a slower innovation clock |
| SG&A | 4.0 | 8.9 | +4.9 | ASML sells 327 lithography systems to four customers who are 61.2% of revenue; KLA serves 57,000 tools at 4,000 facilities with 1,600 to 1,700 applications engineers |
| Gross margin | 51.8 | 60.7 | +8.9 | The mirror of COGS: vertical position, not pricing |
| EBIT margin, ex-impairment | 33.1 | 40.0 | +6.9 | Reconciles exactly: plus 8.9 gross margin, plus 2.8 lower R&D, less 4.9 higher SG&A |
| ASML, % of net sales | COGS | R&D | SG&A | Gross margin | EBIT margin |
|---|---|---|---|---|---|
| FY2023 (calendar) | 48.7 | 14.4 | 4.0 | 51.3 | 32.8 |
| FY2024 (calendar) | 48.7 | 15.2 | 4.1 | 51.3 | 31.9 |
| FY2025 (calendar) | 47.2 | 14.4 | 3.9 | 52.8 | 34.6 |
| KLA, % of total revenues | COGS | R&D | SG&A | Gross margin | EBIT incl. impairment | EBIT ex-impairment |
|---|---|---|---|---|---|---|
| FY2024 (to 30 Jun 2024) | 40.0 | 13.0 | 9.9 | 60.0 | 34.1 | 37.1 |
| FY2025 (to 30 Jun 2025) | 39.1 | 11.2 | 8.5 | 60.9 | 39.3 | 41.2 |
| FY2026 (to 30 Jun 2026) | 38.7 | 11.3 | 8.3 | 61.3 | 41.7 | 41.7 |
Raw values. ASML (EUR m): FY2023 sales 27,558.5 / COGS 13,422.4 / R&D 3,980.6 / SG&A 1,113.2 / EBIT 9,042.3; FY2024 28,262.9 / 13,770.9 / 4,303.7 / 1,165.7 / 9,022.6; FY2025 32,667.3 / 15,409.3 / 4,698.8 / 1,257.8 / 11,301.4. KLA (USD k): FY2024 revenues 9,812,247 / 3,928,073 / 1,278,981 / 969,509, impairment 289,474; FY2025 12,156,162 / 4,751,867 / 1,360,334 / 1,029,734, impairment 239,100; FY2026 13,579,476 / 5,255,060 / 1,532,118 / 1,131,518, impairment nil.
The most persistent difference is gross margin, and it is present in all three years with no crossover: ASML at 51.3, 51.3 and 52.8 against KLA at 60.0, 60.9 and 61.3, a gap that never falls below 7.2 points and averages 8.9. It is not a pricing gap, which is the intuitive reading and the wrong one, because ASML holds the better price lever of the two and still loses. It is a vertical-position gap. ASML describes itself as an architect and integrator: it designs and assembles, and it buys the physics. EUR 4,406.9m of its FY2025 cost of sales, 28.6%, went to Carl Zeiss SMT alone (inferred: 4,406.9 divided by 15,409.3, both disclosed), under a contract whose price is set by "a variable pricing model determined by the annual financial performance of both ASML and Carl Zeiss SMT GmbH." KLA makes the equivalent hard parts itself. Strategically that means ASML's monopoly rent is contractually shared with a supplier holding a reciprocal monopoly over it, on a formula that rises with ASML's own profitability, so the better ASML does the more of the improvement leaves the building.
The cross-check runs both ways, and it contradicts the naive power map. A company with an absolute monopoly ought to show it in gross margin; ASML does not, and trails the company with 56.5% share by nine points. That is a finding, not a rounding error, because it says the usual moat story - ASML has no competitors, therefore ASML captures the economics - fails at the second step. The confirmation runs the other way for KLA: Wallace's claim that "we constrain market share by gross margin" is exactly what a 60.7% three-year average with a rising trend looks like, corroborated by the segment table, where Semiconductor Process Control earned USD 5,482.8m of segment profit on USD 12,244.7m of revenue in FY2026, a 44.8% segment margin. The one place the cost engine argues for ASML is SG&A, where 4.0% against 8.9% is a real and durable advantage of selling very few very large machines to very few customers - and that same fact is what makes ASML's customer concentration severe. The cost advantage and the concentration risk are one mechanism seen from two sides. Forward, the gap narrows without closing: ASML guided FY2026 gross margin to 54 to 56% on 15 July 2026, up from 51 to 53% in April, as EUV mix rises and EXE dilution fades, while KLA guided the September 2026 quarter to 61.6% and holds a 2030 target of 63.5% plus or minus 50 basis points. On both companies' own numbers the gap is still six to eight points at the end of the decade.
| Metric | Threshold | By when | If it hits, it favors | Where published |
|---|---|---|---|---|
| ASML full-year gross margin | 56% or above | FY2027 results, about January 2028 | ASML | ASML Q4 release and 20-F |
| Low-NA EUV systems recognised in calendar 2027 | 85 or more confirms the plus 30%; 78 or fewer means Zeiss capped it | FY2027 20-F, about February 2028 | 85+ ASML; 78 or fewer KLA | ASML 20-F Note 2 unit table |
| KLA gross margin, two consecutive quarters | 63% or above favors KLA; 60% or below favors ASML | any point through June 2028 | either | KLA quarterly releases |
| KLA services revenue growth | 15% or more year on year, the top of a 13 to 15% target Higgins called "conservative" on 9 Sep 2026 | FY2027 10-K, August 2027 | KLA | KLA 10-K Note 17 |
| ASML China revenue as a share of total | 22% or below, from 29.1%, after the BIS Affiliates Rule is reimposed on 10 Nov 2026 | FY2026 20-F, about February 2027 | KLA | ASML 20-F geographic note |