ASML Holding N.V.: A €43–45bn Guidance Reset, 38x Forward Earnings, and No Conservative Margin of Safety

Customer evidence broadly corroborates the story. TSMC raised its 2026 capital-spending plan to $60–64bn, or approximately €51.9–55.4bn at the August 10 ECB rate, from $52–56bn earlier in the year, citing structural demand including AI. Micron has moved toward roughly $27bn of fiscal-2026 capex, about €23.4bn at the same FX rate, and has said equipment investment should rise further in fiscal 2027. SK hynix reported record Q2 2026 results, began HBM4 mass shipments and continues to lay out very large multi-year DRAM/HBM capacity projects. Intel is already using High-NA EUV on selected Intel 18A products. The evidence spans enough customers that ASML’s H1 order strength cannot be dismissed as a single-customer event.

None of that means the semiconductor-equipment cycle has disappeared. ASML’s own recent history argues otherwise. Revenue jumped 30% in 2023, barely grew in 2024, then accelerated 15.6% in 2025 before the current 2026 step-up. In 2025 management explicitly described mainstream, non-AI semiconductor demand as weak for much of the year while AI-related demand strengthened. A secular increase in computing intensity can raise the trend line without eliminating inventory cycles, customer digestion periods or fab-capex pauses.

My qualitative portrait is “high-quality compounding growth with a powerful capex cycle superimposed on it.” Calling ASML merely a cyclical equipment manufacturer misses the EUV monopoly, installed-base economics and rising lithography intensity. Calling it non-cyclical would be equally wrong. The business runs into capex pauses, customer timing changes and geopolitical shocks again and again; the premium valuation ensures those pauses matter to shareholders.

The largest analytical change since the prior report is also the largest new source of uncertainty: ASML has stopped publishing quarterly net bookings. The last hard quarterly figure was Q4 2025, when bookings reached €13.2bn, including €7.4bn of EUV, and year-end backlog stood at €38.8bn. ASML has said quarterly booking figures were volatile enough to mislead investors, and from 2026 the market must rely more heavily on backlog, customer discussions and guidance. Management now says H1 order momentum was “extremely strong,” that backlog has continued to increase, that 2027 low-NA EUV capacity is close to fully covered by orders, and that a significant number of 2028 low-NA orders are already in hand.

Those statements are meaningful substitutes, but they are inferior to a numeric quarterly order series for detecting inflections. Investors can no longer calculate quarterly book-to-bill, see the precise EUV/DUV order mix, or identify whether an abrupt change is broad or customer-specific. So I widen the uncertainty around my 2027 revenue estimate. With quarterly bookings I would normally treat a ±5% range around a well-covered one-year semiconductor-equipment forecast as reasonable; without them I would use roughly ±8%. The wider band is my analytical judgment. It does not erase the visibility supplied by near-full 2027 EUV order coverage, customer capex and annual backlog, but it costs confidence at the margin.

High-NA deserves similar restraint. The technology has crossed a genuine threshold because Intel is using High-NA EUV on selected Intel 18A layers in production. Yet this remains early deployment. ASML recognized three EXE High-NA systems in H1 2026 for €1.189bn of system revenue, versus 29 low-NA NXE systems for €6.709bn. The implied recognized revenue per EXE was about €396m, versus €231m per NXE, roughly a 71% premium. These are revenue-recognition averages, not pure contractual ASPs, so options, configuration and acceptance timing matter. ASML’s 2025 annual report explicitly said EXE systems were dilutive to gross margin, and the company has not disclosed a standalone 2026 EXE gross margin.

ASML reported as early as 2022 that all of its then-current EUV customers had placed High-NA orders, so customer commitment extends beyond Intel. Nobody has yet publicly established broad high-volume manufacturing deployment across that customer base. Intel is the public production proof point; the others are still customer roadmaps and ordered systems, not equivalent HVM evidence.

My base case gives High-NA only a supporting role. I assume roughly €2.5–3.5bn of 2027 revenue from High-NA, about 5–7% of my €52bn base sales estimate. Low-NA EUV, DUV immersion and installed-base management provide the great majority of the earnings expansion. A one-year High-NA delay would hurt the valuation narrative and mix, but it would not by itself break my base business forecast.

China is another part of the story that looks different once you treat it as a path instead of a snapshot. China represented roughly 29% of 2025 sales, all from technologies other than EUV because EUV exports are restricted; ASML now expects approximately 20% of 2026 sales from China. At the €44bn midpoint that equals about €8.8bn. Management says incremental Chinese demand is primarily mainstream Logic rather than leading-edge AI logic. That means China revenue carries a different structural quality from the TSMC/Samsung/Intel advanced-node opportunity.

The Dutch government has progressively expanded licensing requirements for advanced immersion lithography, and EU controls subsequently incorporated some Dutch controls. A proposed U.S. MATCH Act would go further if enacted and internationally implemented, potentially curtailing older equipment and service to additional Chinese customers. As of the research cut-off it remained a proposal rather than an operative blanket ban. Reuters reported one analyst estimate that a severe additional restriction could reduce ASML EPS by up to 10%.

My base case assumes China gradually falls toward a mid-teens percentage of revenue as non-China advanced Logic and Memory grow faster. If China fell to 10% of a hypothetical €52bn 2027 sales base, revenue there would be only €5.2bn, €3.6bn below the 2026 midpoint implied by a 20% share. A portion could be replaced by capacity outside China because global wafer demand does not disappear merely because production geography moves, but replacement is neither instantaneous nor guaranteed. A sustained China reduction without offsetting non-China capacity would take several euros off EPS.

The market faces a cleaner bull/bear disagreement than the “great company versus expensive stock” shorthand suggests. Bulls see the Q2 guidance reset, close-to-full 2027 EUV order coverage, customer capex and 2027–2028 capacity plans as evidence that 2026 is the start of another earnings stair-step. Bears see a €582bn company capitalizing that stair-step before the customer capex cycle has proven durable, with quarterly order data now unavailable and a substantial China business facing political risk.

The prior internal report’s valuation argument has weakened. It has not disappeared.

Vertical history and financial review

ASML began in 1984 as ASM Lithography, a joint venture between Philips and ASM International created to commercialize Philips’ PAS 2000 wafer stepper. The early company lacked the financial scale and market position enjoyed by Japanese lithography leaders. It consumed cash, struggled to establish a customer base and came close to losing parental support. The PAS 5500 family changed the trajectory in the early 1990s. It established a modular platform that could be upgraded over successive process generations. That platform brought major customers into the franchise and gave ASML enough economic footing to finance more aggressive technology development. ASML became an independent public company in 1995, listing in Amsterdam and New York.

The historical IPO archive confirms the 1995 listing, while a semiconductor-industry history records an offering of 12.6m shares in March 1995. I did not recover a primary prospectus line giving the original offer price and exact primary-versus-secondary proceeds during this research pass, so I leave those figures unstated rather than import a secondary estimate into the valuation history.

The next stage turned mechanical precision into a manufacturing architecture advantage. TWINSCAN, introduced around the turn of the century, allowed wafer measurement and exposure tasks to overlap through dual stages, improving throughput. ASML then took the leading position in immersion lithography during the 2000s. The importance of these developments was cumulative: each generation enlarged the installed base, deepened customer process integration and provided engineering learning that could be reused in the next platform. The business model evolved from selling discrete steppers toward selling an upgradeable lithography platform plus lifetime service.

EUV was the fate-changing bet. Producing 13.5-nanometer EUV light with usable source power, reflecting it through defect-free multilayer mirrors, controlling contamination and moving wafers at production throughput required a systems-engineering project spanning optics, light sources, mechatronics, masks, resist and process software. ASML reduced some of the coordination risk by bringing critical capabilities closer. Its 2012 customer co-investment program brought Intel, TSMC and Samsung into a five-year €1.38bn R&D commitment aimed at accelerating EUV and 450mm technology; Intel and Samsung also bought ASML shares, with TSMC subsequently participating. The same period included ASML’s move to acquire Cymer, the critical EUV light-source supplier.

That episode mattered beyond financing. Customers with competing semiconductor roadmaps effectively funded the same lithography platform because none could afford for EUV to fail. It is one of the clearest pieces of historical evidence for ASML’s unusual industry position: the buyers themselves helped finance the supplier whose future pricing power they would later face.

By the late 2010s and early 2020s, EUV moved from technological possibility to high-volume manufacturing. The capital market correspondingly changed its label for ASML. A business once valued as a cyclical semiconductor-equipment supplier increasingly became valued as a monopoly-like technology franchise. The share-price effect was dramatic: ASML’s US line rose sharply through the 2019–2021 EUV adoption phase, fell about 31% in 2022 as rates rose and the semiconductor cycle softened, recovered in 2023, and then moved through another volatile period as AI optimism, China controls and customer spending timing competed for attention. The US listing finished 2025 at $1,064.74 and reached an all-time closing high of $1,986.87 on June 30, 2026 before retreating; the Amsterdam line was €1,513.80 on August 10.

One event from the latest cycle illustrates why ASML can never be valued solely as a smooth compounder. In July 2025 the shares fell sharply after management said it could not yet confirm growth in 2026 despite strong AI demand. Orders and customer timing matter because an EUV machine is an enormous capital item and a handful of customer schedule changes can move quarterly numbers. The subsequent record €13.2bn of Q4 2025 bookings and successive 2026 guidance increases flipped that narrative.

The financial record captures the same mixture of secular growth and lumpiness:

The financial statements show why annual growth rates need interpretation. 2022 was distorted by supply constraints and ASML’s fast-shipment practice, which deferred customer acceptance and revenue; 2023 then benefited as supply caught up, with 53 EUV systems recognized versus 40 in 2022 and a large increase in DUV volumes. 2024 revenue was almost flat as customers digested capacity, while service and field-option revenue still grew 15.5%. In 2025, system revenue increased 12.4% and installed-base revenue 26.2%, producing the highest net income in company history at that point.

The margin record is equally revealing. ASML has operated around a low-50s gross margin through very different semiconductor conditions. That stability comes from pricing, product progression and the growing installed base, but product ramps can temporarily work in the opposite direction. The 2025 annual report specifically identifies early EXE High-NA systems as gross-margin dilutive. The move from 51.3% gross margin in 2024 to 52.8% in 2025 came instead from favorable low-NA NXE mix and higher service/field-option revenue and margins.

2026 marks another step. Q1 margin was 53.0%; Q2 reached 54.0%; Q3 is guided to 55–57%. Management attributed much of the Q2 upside to approximately €300m more installed-base revenue than expected, particularly very high-margin components. That is an important quality distinction. The current margin expansion is being helped by mature installed-base economics and low-NA productivity upgrades alongside the high-cost High-NA ramp.

The cash-flow record is healthy over a cycle even though any individual half-year can look ugly. From 2021 through 2025, aggregate operating cash flow was about €48.6bn against €36.5bn of aggregate net income, an OCF/net-income ratio of approximately 1.33x. The annual ratios were highly volatile, ranging from about 0.69x in 2023 to 1.84x in 2021, largely because customer prepayments, receivables, inventories and system acceptance can shift billions between periods.

That timing issue is visible again in 2026. H1 operating cash flow was negative despite more than €5bn of profit, driven by working-capital movements; Q2 by itself returned to €1.70bn of operating cash flow and about €1.32bn of free cash flow. I do not read the H1 cash-flow deficit as a deterioration in earnings quality. The five-year conversion record carries more information than one half-year.

Capital intensity remains modest relative to the economic value of the machines. ASML generated €12.66bn of operating cash flow in 2025 and spent €1.57bn on property, plant and equipment plus €58m on intangibles, leaving €11.0bn of reported free cash flow. R&D, expensed through the income statement, was far larger at €4.70bn. That accounting structure is important: ASML’s true competitive reinvestment burden sits much more heavily in R&D and supplier engineering than in conventional factory capex.

ASML does not disclose maintenance and growth capex separately. I estimate that roughly €1.0–1.2bn of 2025 capital spending represented maintenance/replacement needs and roughly €0.4–0.6bn growth capacity. This is an analytical estimate, not company disclosure. Because depreciation and maintenance requirements appear broadly similar in magnitude, owner earnings remain close to accounting earnings over time. There is no 30%-plus accounting-to-owner-earnings gap that would force the valuation framework away from earnings.

The balance sheet gives ASML flexibility. At June 28, 2026 it held €6.67bn of cash plus €0.91bn of short-term investments. Long-term debt was about €1.98bn. Inventory was €11.74bn, substantial but understandable for systems whose lead times, components and customer acceptance cycles span many months. Receivables were also large, reinforcing the working-capital volatility.

Capital allocation has been shareholder-friendly without starving engineering. ASML spent €4.7bn on R&D in 2025 and returned €8.5bn to shareholders. It completed its previous buyback after repurchasing €7.6bn of shares and announced a new program of up to €12bn through the end of 2028, with most repurchased shares intended for cancellation. The 2025 dividend was €7.50 per share, and the July 2026 interim dividend was €1.88.

Governance is conventional for a large Dutch public company in most economic respects. Christophe Fouquet became president and CEO in April 2024 after a long internal career, replacing Peter Wennink; Roger Dassen remains CFO, while Marco Pieters joined the Board of Management as CTO in April 2026. No founder or operating shareholder controls the company. Dutch preference-share foundation arrangements provide a potential takeover-defense mechanism, a governance feature worth remembering but not a current source of operating control.

The vertical lesson holds across forty years: ASML’s defining skill has been coordinating technologies that no individual component supplier or semiconductor customer could industrialize alone. PAS 5500 established the manufacturing base. TWINSCAN and immersion won throughput leadership. EUV turned systems engineering into a monopoly position, and installed-base upgrades turned a historically lumpy tool business into something more recurring. High-NA is the next test of the same capability rather than an entirely new business model.

Business model, moat, industry and peers

ASML’s 2025 revenue mix makes the economic machine visible. Net system sales were €24.47bn, or about 75% of revenue, while service and field options were €8.19bn, about 25%. H1 2026 moved further toward the installed base: total sales were €18.1bn, net system sales €12.84bn and service/field options roughly €5.25bn.

Within H1 2026 system revenue, the technology mix was already heavily EUV:

Counting machines is misleading, as the table shows. Thirty-two EUV units generated almost €7.9bn of system revenue in H1, while a much larger number of mature-node and metrology systems contributed much less revenue per unit. It also illustrates why High-NA can matter financially long before its unit volume looks large. The EXE average recognized revenue was roughly 1.7x NXE in H1 2026.

The moat is real because it is systemic rather than patent-deep. The first layer is technology integration. EUV requires an extreme-ultraviolet light source, ultra-flat reflective optics, vacuum systems, nanometer-stage control, metrology, computational correction and enormous software content to work together at production throughput. A rival cannot close the gap by reproducing one optical element or buying one missing supplier.

The second layer is cumulative learning. Semiconductor fabs optimize process flows around actual scanner behavior. Every installed machine creates data on overlay, focus, source performance, resist behavior and throughput. ASML can feed that information into both software upgrades and the next hardware generation. A challenger would begin without decades of production learning at the customers whose process engineers determine whether a new tool can enter volume manufacturing.

The third layer is supplier architecture. ASML coordinates thousands of suppliers and spent €4.7bn on its own R&D in 2025. The earlier decision to acquire Cymer and the longstanding relationship with specialist optics suppliers illustrate how the company has made critical subsystems part of a managed ecosystem rather than treating them as commodity inputs. The 2012 customer co-investment episode showed that customers themselves regarded this ecosystem as strategically irreplaceable.

The fourth layer is switching cost. A lithography platform is incorporated into node development years before volume production. Qualification, masks, process windows and production software are all built around it. In EUV there is currently no alternate supplier to switch to. In DUV, Nikon and Canon retain relevant technology, but neither provides a commercial substitute for ASML EUV. That makes ASML a rare case where customer bargaining power is simultaneously enormous in absolute purchasing scale and weak at the critical technological bottleneck.

The moat does have boundaries. Customers determine when fabs are built, how many wafer starts are required and whether a process uses a particular mix of EUV versus multipatterned DUV. TSMC, Samsung and Intel are sophisticated counterparties capable of delaying or reallocating billions of euros of equipment. ASML controls the critical tool; it does not control the semiconductor cycle.

Industry economics reinforce that position. SEMI expected global semiconductor-manufacturing-equipment sales to rise about 9% in 2026 to roughly $126bn and another 7.3% in 2027, driven by AI-related Logic and Memory. At the August 10 ECB rate those amounts are about €109bn and €117bn. Other industry forecasts published during 2026 have moved even higher as AI infrastructure plans expanded, a sign of how quickly expectations are changing.

The key secular driver is lithography intensity rather than semiconductor sales alone. Advanced Logic nodes need more expensive critical-layer patterning, and DRAM scaling and HBM manufacturing are raising advanced lithography demand while advanced packaging adds another equipment layer on top. That broadens the profit pool for the entire wafer-fab-equipment group. ASML captures the lithography bottleneck, KLA captures process-control intensity, Lam Research benefits from etch/deposition steps, Applied Materials from broad materials engineering, and Tokyo Electron from a diversified set of deposition, etch and coat/develop tools.

ASML therefore falls into the “no directly comparable company” competitive scenario for its core EUV franchise. Broader semiconductor-equipment companies are useful financial comparables because they share the same customers and capex cycle, yet none duplicates ASML’s EUV economics.

Applied Materials became the broadest materials-engineering platform in the group. Customers use it across deposition, materials modification and other front-end steps, which spreads its revenue across more process categories than ASML. That breadth reduces dependence on a single technology roadmap but also means Applied operates in more categories with direct competitors.

Lam Research became especially important where increasingly three-dimensional device structures create more etch and deposition steps. Memory cycles matter heavily here. The AI/HBM boom is attractive to Lam for a different reason from ASML: ASML benefits when finer patterning requires more advanced lithography, while Lam benefits when vertical and structural complexity requires more process steps.

KLA occupies process control. As feature dimensions shrink and process windows tighten, the economic value of detecting a defect before hundreds of subsequent steps rises. That can make inspection and metrology spending relatively resilient and supports unusually high margins. A third-party industry analysis estimates KLA’s process-control share at roughly 50–55%, though that figure is not a company filing and should be treated as directional.

Tokyo Electron is the strongest Japanese broad-line comparison, particularly across coater/developer, deposition and etch. Nikon is more useful as the direct historical lithography comparison. ASML’s history matters here because lithography competition was once genuinely multipolar; EUV turned it into a category in which the direct peer disappeared.

A same-basis valuation snapshot makes the sector-wide enthusiasm obvious. The following figures use trailing P/E for every company, from the same August 2026 third-party data snapshot; they are not mixed with ASML’s forward multiple:

The important point is not that ASML is “cheap versus peers.” It is not. The entire capital-equipment group has been re-rated around an unusually strong AI-capex narrative. ASML receives only a modest trailing premium to Applied Materials and trades close to KLA and below Lam in this snapshot. That relative comparison cannot justify the absolute valuation because peer multiples themselves are elevated.

Forward earnings tell a more company-specific story. Using my FY2026 EPS estimate of €39–40, ASML trades around 38–39x current-year earnings. My base 2027 EPS estimate is €48.7, putting the stock around 31x that number. The premium becomes manageable if the 2027 revenue acceleration occurs. If 2027 disappoints and EPS remains around €40, investors are still paying roughly 38x with little earnings growth to absorb multiple compression.

This is where capital-market expectations and business quality intersect. ASML has earned a premium because its market share in the most valuable lithography category is stronger than the competitive positions of broad WFE peers. The premium remains justified in direction. The open question is magnitude.

The 2024 Investor Day framed 2030 revenue opportunity at €44–60bn and gross margin at 56–60%. With the new 2026 revenue midpoint already €44bn, the low end of that old 2030 revenue framework has effectively arrived four years early. That should not be read as evidence that growth is finished; the framework was scenario-based and predates the current AI capex acceleration. It does mean the old range has lost usefulness as an outer bound. ASML’s June 10, 2027 Capital Markets Day will be a critical opportunity to reset it.

My base case assumes a structural increase in semiconductor capital intensity layered on top of a surviving semiconductor cycle. AI is raising the trend line; it has not repealed capital discipline. This view would be falsified by a coordinated reduction in leading-edge Logic and DRAM investment, falling ASML utilization-driven service demand, weakening 2027 order coverage or a sustained gross-margin reversal despite product maturity.

Current fundamentals and visibility reconstruction

The last four reported quarters show a business that moved from “solid AI-assisted recovery” into genuine acceleration:

Q4’s €13.2bn order intake and €38.8bn year-end backlog were the handoff into 2026. Q1 then raised the annual guide to €36–40bn from January’s €34–39bn starting range. Three months later Q2 moved it again to €43–45bn. A seven-and-a-half-billion-euro shift in midpoint guidance between January and July is large enough that valuation work built on the January earnings base is obsolete.

The gross-margin progression matters just as much. Q2 had originally been guided to 51–52%. Actual gross margin was 54%. Management attributed the upside to the mix of installed-base sales, including high-margin component activity, rather than suggesting that every new High-NA machine suddenly reached mature profitability. Q3’s 55–57% guide takes the business toward the lower end of ASML’s old 2030 margin ambitions years ahead of schedule.

The bookings blackout removes one of the best short-term warning signals, while leaving several slower but useful substitutes. Use them together, not one at a time.

The first substitute is annual backlog. The last hard figure, €38.8bn at December 2025, exceeded one year of 2025 revenue and represented roughly 88% of the new €44bn 2026 midpoint. Backlog is not equivalent to next-year sales because delivery and acceptance can span periods, but it anchors demand already contractually present.

The second substitute is order coverage by product. ASML says 2027 low-NA EUV capacity is close to fully covered and that it has already received a significant volume of 2028 low-NA orders. That statement provides less precision than a booking number but arguably more direct information about the revenue-critical part of the portfolio. Low-NA EUV carries far more revenue per system than mature DUV, making coverage there particularly valuable.

The third substitute is ASML’s own capacity plan. Roughly 65 low-NA EUV machines in 2026 followed by a 30% increase implies capacity around 85 systems in 2027; another 30% would take potential 2028 capacity to roughly 110. Similar 30% increments from about 130 immersion machines imply the ability to move toward roughly 170 in 2027 if demand and supply support it. These are calculated capacities, not shipment or revenue forecasts.

The fourth substitute is customer capex, where evidence is unusually strong. TSMC’s €51.9–55.4bn equivalent 2026 plan is substantially above 2025 spending. Micron has said fiscal-2027 equipment spending will rise further after a fiscal-2026 plan around €23bn equivalent. SK hynix’s July results describe AI-memory demand, HBM4 shipment and continued capacity expansion. Samsung remains exposed to the same DRAM/HBM and foundry investment cycle. Intel is a weaker financial contributor than TSMC or the Korean memory leaders, but it is strategically valuable because its 18A deployment gives High-NA a production reference.

The fifth substitute is installed-base activity. A >30% annual increase in ASML’s service and field-option business indicates that customers are actively pushing existing scanners for more output. In a real capex collapse, utilization-sensitive upgrades and components would eventually soften. That makes installed-base management (IBM) revenue a useful indirect demand gauge when quarterly bookings are absent.

What remains lost is the speed of detection. An order cancellation or customer pause that occurs in August might not become numerically visible until revenue guidance changes or the next annual backlog is published. Market participants previously had quarterly bookings as an intermediate signal. I attach lower confidence to exact quarterly and one-year estimates than I would have under the old disclosure regime.

High-NA is the other area where confidence needs calibration. ASML’s H1 statutory data show three EXE systems recognized for €1.189bn, up from one system and €274m in H1 2025. Low-NA NXE revenue was €6.71bn from 29 systems. The platform is becoming material enough to move revenue, but its gross-margin contribution remains undisclosed.

Intel’s production milestone narrows the technological risk. Selected 18A layers have been dual-qualified on High-NA and existing NXE tools, and ASML says yield performance matched the NXE platform in Intel’s Oregon development/manufacturing environment. Intel was the first customer to receive commercial EXE:5000 modules and the first to install and accept the production-oriented EXE:5200B.

High-NA has graduated from laboratory optionality into early production economics, while broad industry deployment remains unproven. My 2027 estimate deliberately does not require a High-NA explosion. The base case uses roughly €3bn High-NA revenue. The conservative case assumes a slower ramp around €1.5–2bn; the optimistic case allows roughly €5bn. Each case still gets most revenue from established low-NA EUV, DUV and the installed base.

China requires the inverse discipline: do not extrapolate the recent peak. China contributed roughly €9.5bn in 2025, around 29% of ASML sales. Management now expects about 20% in 2026, implying €8.6–9.0bn under the current annual guidance. Q2 system-sales geography was already much more weighted to South Korea and Taiwan than China, consistent with advanced Logic and Memory investment taking the growth baton.

The risk from China has two layers. Regulation can directly block equipment and service. Domestic substitution can gradually erode ASML’s mature-node DUV opportunity even where exports remain legal. Reports in July suggested Chinese entities were working toward domestic immersion DUV production, but public evidence on volume, overlay, uptime and customer qualification remains insufficient to treat those systems as an ASML-class competitor. I exclude meaningful Chinese lithography substitution from the 2027 base case and include it as a longer-dated downside.

The market is currently trading the convergence of four real fundamentals: higher 2026 earnings, 2027 low-NA order coverage, leading-edge customer capex and rising IBM revenue. Layered on those fundamentals is a broader narrative that AI infrastructure spending can continue almost without interruption. The first four are observable. The fifth is the assumption whose durability will decide the multiple.

For the next earnings release on October 14, the market should care less about one quarter of revenue than about three pieces of information: whether the €43–45bn FY2026 guide rises again or at least becomes more secure; whether management repeats the 2027/2028 EUV coverage and capacity language; and whether the 55–57% Q3 gross-margin range proves repeatable rather than mix-driven.

Valuation, risks, catalysts and tracking

The first valuation step is cash-flow passthrough. ASML passes it. Aggregate operating cash flow from 2021–2025 was approximately €48.6bn against €36.5bn of net income, a ratio of 1.33x. Reported 2025 free cash flow was €11.0bn versus €9.6bn net income. Working capital makes individual years volatile, but accounting profit has converted into cash over the cycle.

Using the approximate €582bn current market value, 2025 FCF implies only a 1.9% trailing FCF yield. Normalizing to my roughly €14–15bn FY2026 owner-cash-earnings estimate raises that to about 2.4–2.6%. The Netherlands 10-year government bond yielded approximately 3.28% on August 10. ASML therefore offers a lower current cash yield than the sovereign risk-free alternative; investors are explicitly paying for growth.

Maintenance capex does not materially change this conclusion. ASML spent €1.63bn on PP&E and intangibles in 2025, and I estimate €1.0–1.2bn of that as replacement/maintenance spending. Since the estimated maintenance burden is close to depreciation scale, owner earnings remain near net income. The difference between owner-earnings and headline P/E is comfortably below the 30% threshold at which I would abandon earnings-based valuation. The valuation uses forward earnings, with FCF/DCF as a cross-check.

Historical P/E is a difficult comparison because 2026 earnings are rising much faster than trailing earnings. One third-party historical series places ASML’s recent trailing P/E well above its 3- and 5-year averages near 37–38x, while another forward series places the current forward ratio around 39x. I give more weight to the guidance-derived forward multiple because the denominator reset is so large.

At €1,513.80, my FY2026 midpoint EPS near €39.4 produces a current-year forward P/E of about 38.4x. My 2027 base EPS estimate of €48.7 lowers it to about 31.1x. That is still a premium multiple, but it is a very different proposition from paying 50–60x for flat earnings.

My scenario framework is deliberately wider than the old report because the bookings blackout reduces precision:

The conservative case assumes AI capex remains strong enough to avoid a semiconductor recession but 2027 machine capacity is under-utilized relative to the current plan, China falls materially and High-NA ramps slowly. The base case assumes ASML converts most of its disclosed 2027 low-NA capacity expansion into revenue, installed-base sales keep growing and the gross-margin structure begins moving toward the old long-term range. The optimistic case requires both advanced Logic and Memory capex to stay unusually strong, High-NA production to broaden and service/upgrades to keep supporting mix.

This is valuation-scenario analysis within a research framework, not investment advice.

A DCF sanity check reaches a similar conclusion only under demanding assumptions. Starting normalized owner cash flow around €14–15bn, a base DCF needs double-digit annual cash-flow growth through much of the next five years, a EUR discount rate around the mid-7% range and a terminal growth assumption in the mid-3% range to support roughly €1,500 per share. Raise the discount rate toward 9% and slow the first five years materially, and value can fall below €1,000. ASML’s moat makes a long runway plausible; the current price already monetizes much of that plausibility.

The most fragile base-case assumption is conversion, not High-NA: turning 2027 customer-capacity plans into delivered, accepted and paid-for systems at the expected gross margin. If only 70% of the base case’s incremental 2027 sales growth above €44bn materializes, I get approximately €49.6bn revenue. With gross margin around 55.5%, estimated EPS falls toward €45.3. Applying the same 30–33x base multiple gives roughly €1,358–1,493, midpoint around €1,426. That is below the current price.

The flat-earnings test is harsher. Suppose ASML earns roughly €39.4 per share in 2026 and earnings do not grow for the subsequent three years. With no multiple compression, the shareholder’s return would essentially consist of the dividend, well below the 3.28% Dutch 10-year yield. At a still-premium 30x exit P/E, €39.4 of earnings would imply a share price around €1,182 before dividends, producing an approximate three-year annualized loss of about 8%. There is no margin of safety at this buy price under a flat-earnings outcome.

Margin-of-safety verdict: none. The current price is approximately 36–52% above my conservative-value band. The business can still compound successfully from here; conservative downside simply offers no valuation cushion if the cycle stops cooperating.

For a 3–5-year investor, my rough total-return assumptions are wider than the 12-month fair values. Under a conservative path with 2029 EPS only around €42 and a mid-20s exit multiple, annualized return could be roughly -10% to -13%. A base path with 2029 EPS around €60–65 and roughly 30x valuation produces approximately 6–9% annualized. An optimistic path with EPS approaching €75–80 and a low/mid-30s multiple can produce roughly 18–21% annualized returns. Those figures make the current quote look acceptable for existing holders who strongly believe in the base-to-bull growth path, rather than compelling for a new investor requiring conservative downside protection.

The permanent-loss risks are more specific than ordinary semiconductor volatility.

The first is an AI-capex digestion cycle. Probability: medium. Impact: high. TSMC, the three DRAM suppliers and hyperscalers have simultaneously raised investment. That is exactly the environment in which equipment lead times and capacity expansion can overshoot. A 2027–2028 customer capex pause would hit new systems first, then upgrades/utilization. Revenue growth would disappear while ASML’s own R&D and engineering base remained largely fixed; EPS would fall faster than revenue, and the multiple could revert toward a conventional cyclical-equipment range. The indicators are customer capex cuts, weaker low-NA order coverage, slower IBM growth and ASML reducing its planned 2027–2028 manufacturing capacity.

The second is China regulation. Probability: medium. Impact: medium to high. A more comprehensive ban on immersion DUV and service to major Chinese fabs could remove several billion euros of annual business before non-China capacity is ready to replace it. The transmission path is direct sales loss, weaker factory absorption, then lower EPS and a narrative shift from “global lithography monopoly” toward “politically segmented addressable market.” The observable indicators are Dutch/EU license changes, enactment and allied implementation of U.S. legislation, or ASML reducing its approximately 20% China expectation.

The third is High-NA economics. Probability: low to medium; long-term impact high. A delay would matter less to 2027 revenue than the market narrative implies because my base case is not High-NA-heavy. A persistent failure to achieve customer productivity, availability or economically acceptable gross margin would matter much more. The observable indicators are EXE unit recognitions, additional disclosed HVM customers, customer node decisions and whether ASML stops describing EXE as margin dilutive.

The fourth is valuation compression. Probability: medium to high; impact high. A 38–39x current-year P/E offers little protection against even moderate earnings disappointment. The Dutch 10-year yield is already above ASML’s normalized current cash yield. If rates rise further or investors conclude that the WFE supercycle is merely a strong cycle, the multiple can fall even while earnings remain positive. A move from 38x to 28x on €40 EPS would take the stock close to €1,120 without an earnings recession.

The fifth is disclosure risk rather than economic risk. Probability of forecast error: high; business impact: low by itself. Without quarterly bookings, investors will discover some demand inflections later. That raises the chance that consensus estimates remain stale for longer and then adjust abruptly. It deserves a wider scenario range, not a lower moat score.

Positive catalysts over the next year include another FY2026 guide increase, evidence that 2027 EUV capacity remains fully ordered, further TSMC/memory capex increases, additional High-NA production customers and gross margin moving toward 57–58% without dependence on one-off installed-base components. The 2027 Capital Markets Day could become the largest medium-term catalyst because the old €44–60bn 2030 revenue framework is now too low at its bottom end to describe the changed earnings base.

Negative catalysts are the mirror image: a guide cut, customer schedule push-outs, a gross-margin reversal, tougher China controls, evidence that 2027 capacity was built ahead of demand, or High-NA acceptance delays.

The practical tracking dashboard is:

ASML itself is the primary source for revenue, margin, IBM, capacity and earnings-date tracking. Customer IR materials matter for capex. Dutch/EU/U.S. government releases matter for China controls. The bond market and share-price multiple should be treated as valuation context rather than business indicators.

Cross-synthesis and final conclusion

ASML has proven one capability more convincingly than anything else across its full history: it can industrialize technology that sits near the physical limit of what semiconductor manufacturers require, while coordinating customers and suppliers through development cycles that run for a decade or more.

That capability is harder to copy than an individual machine specification. Japanese lithography competition once looked formidable. ASML did not win through a single lucky product. PAS 5500 provided a platform, TWINSCAN improved manufacturing economics, immersion established process leadership, and EUV forced the company to coordinate light-source physics, optics, vacuum engineering, mechatronics, software and customer process integration. The customer co-investment program, Cymer integration and subsequent EUV ramp all expressed the same institutional ability.

Those success factors remain present. R&D spending continues to rise, reaching €4.7bn in 2025, and the installed base is both larger and generating €8.2bn of annual service and field-option revenue. Customers are financially committing to another round of advanced-node and memory capacity. High-NA has reached production use rather than remaining a development demonstrator.

The weakness is that this exceptional competitive structure exists inside an industry where customers still control the capital-spending clock. ASML’s monopoly cannot make TSMC build a fab it does not need. It cannot force DRAM suppliers to add capacity during an inventory correction. It cannot eliminate U.S.-China policy. It cannot prevent a €400m scanner from slipping between quarters because customer acceptance moved.

That distinction explains the valuation problem better than the word “expensive.” A monopoly supplier with recurrent installed-base revenue deserves a premium to a generic machinery manufacturer. A company whose forward EPS may rise from €24.73 in 2025 to roughly €39–40 in 2026 deserves to be valued on forward rather than stale trailing earnings. Yet a cyclical business at roughly 38–39x current-year earnings still requires unusually durable growth.

The market may currently be misjudging two things in opposite directions.

It may still underestimate how dramatically the near-term earnings base has shifted. Q2’s guide increase means the old static P/E objection has lost much of its force. At my base 2027 EPS, the multiple is close to 31x, a level that is defensible for a company with ASML’s moat if growth continues.

At the same time, the market may overestimate how certain that 2027 number is. Quarterly bookings have disappeared precisely as ASML and its customers plan very large capacity increases. Management’s coverage commentary is strong enough to support the direction, but the missing data make it harder to detect the first downturn. A €582bn valuation leaves little room for that uncertainty.

The one-year debate turns on conversion. ASML already has the demand commentary, capacity plan and customer capex. The next twelve months must show that those plans convert into €50bn-plus 2027 economics rather than remaining capacity capability.

The three-year debate turns on industry structure. My base case assumes AI makes leading-edge logic and memory structurally more capital intensive while semiconductor cycles continue around that higher trend. A temporary 2027 digestion period would not break the thesis. A broad reduction in advanced-node and HBM capital intensity would.

The five-year debate turns on High-NA and the next lithography roadmap. Intel has proved that High-NA can enter production. ASML still needs to prove that the economics work across TSMC, Samsung and memory customers at meaningful scale, and that EXE margins mature rather than remaining diluted indefinitely. A successful High-NA franchise can raise revenue per system materially while extending ASML’s technology lead. A delayed ramp would leave low-NA EUV capable of carrying the business for some time, but it would reduce the long-term growth ceiling.

China is unlikely to determine ASML’s ultimate technological position, yet it can materially alter the earnings path. The 2026 expectation of approximately 20% revenue exposure remains large enough that a regulatory shock could create a multi-billion-euro hole. Over five years, I would prefer to see China fall as a percentage of revenue because non-China EUV and Memory sales grew faster, rather than because export controls destroyed profitable DUV business.

The bookings change costs analytical confidence rather than business quality. Annual backlog, capacity commitments and customer capex are good substitutes for estimating the broad direction. None lets an outside investor reconstruct a quarterly order curve. That deserves a wider discount between an optimistic operational forecast and the price an investor should willingly pay.

The current quote looks like a fair price for the base case and a poor price for the conservative case. This is the material change from the old price-only objection. The raised earnings base has brought the forward valuation down enough that the shares no longer require an obviously extreme multiple if 2027 executes. The absence of conservative downside protection prevents me from moving from “fairly valued high-quality business” to “attractive purchase.”

Bull reasons:

• 2026 revenue guidance rose from €34–39bn in January to €43–45bn by July, while gross-margin guidance moved to 54–56%, establishing a much higher earnings base.
• Management says 2027 low-NA EUV capacity is close to fully covered by orders, with significant 2028 orders already received.
• TSMC, Micron and SK hynix customer evidence independently supports a large leading-edge Logic and Memory capex cycle.
• Installed-base revenue grew 26.2% in 2025 and is expected to rise more than 30% in 2026, adding a recurring, high-margin layer to the system business.
• High-NA has entered production at Intel, while historical order disclosures show commitment extending beyond a single EUV customer.

Bear reasons:

• At €1,513.80 the stock still trades near 38–39x estimated FY2026 earnings and offers a normalized cash yield below the 3.28% Dutch 10-year bond yield.
• Quarterly bookings are no longer disclosed, removing the cleanest early-warning indicator for a business whose revenue can shift with a handful of customer decisions.
• China remains roughly 20% of expected 2026 sales and faces a live risk of tighter DUV/service restrictions.
• High-NA revenue per recognized system is already very large while the company’s latest explicit annual disclosure still describes EXE as gross-margin dilutive.
• The broader semiconductor-equipment peer group is itself trading at elevated multiples, so relative valuation provides little protection if the entire AI-capex complex de-rates.

Pre-mortem, first script: during 2027 the current AI equipment cycle finally outruns wafer demand. TSMC trims its next capex budget, DRAM suppliers postpone part of their HBM/DRAM equipment schedule, and ASML quietly reduces the planned 30% 2028 capacity increment. Because quarterly bookings are unavailable, the downgrade first appears through weaker guidance. Revenue settles near €40–43bn rather than rising above €50bn, gross margin drops toward 50–51%, and EPS falls toward €30–32. A 22–24x cyclical P/E would value the shares around €660–770, roughly 49–56% below the present quote.

Second script: by 2028 tighter allied export controls cut Chinese immersion sales and service faster than non-China mature capacity replaces them, while reported Chinese immersion-DUV alternatives become commercially usable at domestic fabs. High-NA adoption outside Intel also runs a year behind expectations and remains margin-dilutive. ASML still owns EUV, but sales stagnate in the mid-€40bn range, gross margin sits around 51–52% and EPS reaches only €33–35. At 24–26x earnings the stock would be worth roughly €800–900, a loss around 41–47%. The moat survives in this script; the investment fails because earnings and multiple contract together.

Research uncertainties are material in four places. Quarterly bookings and numerical backlog will remain unavailable until annual disclosures, limiting short-cycle visibility. ASML does not disclose EXE-specific gross margin, so High-NA profitability must be inferred from company comments. Public customer announcements do not reveal exact scanner delivery schedules, making a capacity plan different from a revenue forecast. Finally, current peer forward-P/E services use inconsistent fiscal-year and NTM definitions; I therefore used trailing multiples on a common basis for peer comparison and my own guidance-derived forward earnings for ASML.

The source hierarchy behind this report is ASML’s 2025 annual report, Q1/Q2 2026 primary releases, statutory interim statements and earnings-call materials; semiconductor-customer investor relations; Dutch government and ECB data; and Reuters/major financial press chiefly for legislation, market reaction and cross-checks. Secondary valuation databases were used only for directional historical/peer comparisons, never as the primary earnings forecast.

The business conclusion is straightforward. ASML remains one of the hardest semiconductor franchises to displace. The 2026 guidance reset, installed-base acceleration and 2027 EUV coverage materially improve the forward earnings case compared with a valuation analysis built on 2025 static earnings. The current share price can be defended on a 2027 base case without assuming that High-NA immediately takes over the industry.

The investment conclusion is more restrained. At €1,513.80, the shares already sit close to my base-case value. The conservative case is €996–1,116, and the flat-earnings test gives an unattractive outcome against a 3.28% Dutch government yield. A new buyer receives business quality and growth visibility, but little protection if AI capital spending merely pauses. Existing holders have a more reasonable case because selling a scarce franchise solely on headline trailing P/E would ignore the earnings reset.

【Company-profile scores】

• Fundamental quality: high
• Growth: high
• Moat: strong
• Financial soundness: strong
• Management credibility: high
• Valuation attractiveness: medium
• Risk level: medium
• Suitable investor type: long-term growth

【Investment rating】

• Rating: Hold
• One-line thesis: Raised 2026 earnings and near-full 2027 EUV coverage support the valuation, but the current price offers no conservative margin of safety.

【Ideal Buy Price】800–890 EUR

Basis: my conservative 2027 fair-value range is approximately €996–1,116; applying a 20% discount to each end of that range produces roughly €797–893, rounded to €800–890.

• Acceptable hold price: €1,400–1,650, centered on the €1,461–1,608 base-case valuation.
• Clearly overvalued price: €2,250 and above, roughly 10% above the €2,044 midpoint of the optimistic valuation; I use €2,250–2,500 as the overvaluation band.
• Current-price classification: acceptable hold.
• Whether to wait for a better price: yes for a new position requiring a margin of safety. The preferred entry is €800–890; a less stringent growth investor could reassess around €1,100–1,200 if 2027 EUV coverage, customer capex and 55%-plus gross margin remain intact. The opportunity cost is missing another earnings-led re-rating if 2027 revenue exceeds €52bn.
• Target holding horizon: 3–5 years.
• Expected annualized return: conservative approximately -10% to -13%; base approximately 6–9%; optimistic approximately 18–21%, using scenario-specific 2029 EPS and terminal multiples rather than assuming the current multiple persists.
• Max-loss risk: roughly 49–56% in the pre-mortem where the AI equipment cycle rolls over, gross margin returns to about 50–51%, EPS falls toward €30–32 and the P/E compresses into the low-20s.
• Reassessment-trigger signals: FY2026 revenue guide below €43bn; gross margin below 53% for two consecutive quarters; withdrawal of the roughly +30% 2027 low-NA EUV capacity plan; TSMC or major DRAM customers cutting equipment budgets by more than 10%; China restrictions that materially reduce the expected approximately 20% revenue contribution without offsetting non-China demand; High-NA remaining margin-dilutive into broad 2028 volume deployment.

【Valuation Range】

• current: 1,513.80 EUR (close as of 2026-08-10)
• bear (conservative · ideal buy zone): [800, 890]
• base (fair · acceptable hold zone): [1,400, 1,650]
• bull (optimistic · above the clearly-overvalued line): [2,250, 2,500]

Other tickers mentioned

• TSM.US — TSMC is the most important leading-edge foundry capex reference and a central driver of ASML EUV demand.
• INTC.US — Intel provides the first disclosed High-NA production-use proof point through selected Intel 18A products.
• 005930.KO — Samsung Electronics is a major DRAM, foundry and EUV customer whose spending affects both Logic and Memory demand.
• 000660.KO — SK hynix is a major HBM/DRAM investor and an important indicator of advanced-memory lithography demand.
• MU.US — Micron’s rising equipment budget provides an independent cross-check on ASML’s Memory demand outlook.
• AMAT.US — Applied Materials is a broad wafer-fab-equipment valuation and cycle peer.
• LRCX.US — Lam Research is an etch/deposition peer with substantial exposure to advanced Logic and Memory investment.
• KLAC.US — KLA is the process-control peer whose high-margin franchise offers another scarce-equipment valuation reference.
• 8035.TSE — Tokyo Electron is a major Japanese broad-line wafer-fab-equipment peer.
• 7731.TSE — Nikon is ASML’s principal listed historical lithography competitor in DUV, without a commercial EUV alternative.
• 7751.TSE — Canon remains active in lithography and alternative patterning technologies but does not presently substitute for ASML EUV.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.