ASML Holding (ASML): China’s DUV Lithography and the Semiconductor Equipment COMAC Problem

August 6, 2026

Overview

In July 2026, a report that China had begun limited production of a domestically developed immersion deep-ultraviolet lithography system created two opposite reactions. One side treated the news as evidence that China was finally approaching ASML. The other dismissed it by comparing a reported production plan of roughly five tools in 2026 and twenty in 2027 with ASML’s planned 2026 immersion capacity of about 130 systems. Both reactions were too simple.

The first Chinese immersion DUV systems do not represent an immediate substitute for ASML’s most advanced equipment. Their production scale is small, their throughput and overlay performance have not been publicly verified, and their long-term reliability inside high-volume fabs remains unknown. China still does not possess a commercially proven extreme-ultraviolet system. ASML, meanwhile, has moved beyond conventional EUV into High-NA EUV, and in July 2026 announced that Intel Foundry had used the technology for selected Intel 18A production layers. The technological distance between a newly introduced domestic immersion scanner and ASML’s most advanced platform therefore remains enormous.

But the strategic meaning of the Chinese effort is not captured by asking whether it can defeat ASML in the global market. A substitute does not need to become globally competitive before it begins changing the structure of dependence inside its home market. If a state-backed tool can enter SMIC, Hua Hong Semiconductor or ChangXin Memory Technologies, survive real production conditions, receive repeated engineering feedback, and create demand for domestic optics, light sources, stages, metrology, software and spare parts, it can start building an industrial learning system. The tool may be inferior and still be strategically consequential.

This is the semiconductor-equipment version of the COMAC problem. COMAC did not need to displace Airbus or Boeing globally before it changed the future bargaining position of foreign aircraft manufacturers in China. It needed a protected domestic market, state-aligned customers, a patient capital base, and enough operational deployment to accumulate engineering experience. China’s lithography path is harder because a scanner is a much lower-tolerance manufacturing system than a commercial aircraft, but the structural question is similar: can a domestic platform reduce strategic dependence before it reaches the frontier?

For ASML, this creates two different time horizons. In the near term, AI-driven logic and memory investment is expanding demand for EUV, immersion DUV and installed-base upgrades. ASML reported €9.3 billion of second-quarter 2026 revenue, a 54.0% gross margin and €2.9 billion of net income, then raised its full-year outlook to €43–45 billion of sales and a 54–56% gross margin. Its core business is not being displaced. Over the longer term, however, China may cease to be a permanently dependent equipment market even if it remains far behind the frontier. The central issue is not whether ASML still controls the best lithography technology. It does. The issue is whether China can create a lower-performance but self-reinforcing industrial base beneath that frontier.

ASML belongs primarily to the Semiconductors, Compute & Packaging layer of the AI Civilization Map. Export controls and equipment sovereignty bring it into Geopolitics, Sovereignty & Constraints, while customer co-investment and fab capital expenditure connect it to Capital, Institutions & Operating Layers. Those alignments help locate the company, but they do not replace the underlying analysis. This article therefore begins with the physics and industrial history of lithography, then follows the chain into China, High-NA EUV, AI demand, financial structure and valuation.

Why Lithography Controls Semiconductor Manufacturing

A modern semiconductor is produced through hundreds or thousands of process steps, but lithography has a special role because it tells the rest of the factory where to act. Deposition adds material. Etch removes material. Ion implantation changes electrical properties. Cleaning removes contamination. Those steps can be extraordinarily difficult, yet they do not independently determine the geometry of the circuit. Lithography transfers the pattern that defines where the next operation should occur.

The basic cycle resembles photographic printing. A wafer is coated with a light-sensitive material called photoresist. Light is directed through or reflected from a reticle containing the circuit pattern. An optical system reduces and projects that pattern onto the wafer. The resist is developed so that selected areas remain while others are removed. Subsequent process tools then etch, deposit or modify the exposed regions. The cycle repeats layer after layer until a flat silicon disc becomes a three-dimensional electrical structure containing billions of transistors and interconnects.

Three variables make lithography economically decisive. The first is resolution: the system must print sufficiently small features. The second is overlay: every new pattern must align with previous layers at nanometer-scale accuracy. The third is throughput and availability: the machine must process enough wafers per hour and remain operational for enough time to justify its cost. A laboratory demonstration can succeed on resolution while failing commercially because overlay drifts, contamination rises, the stage cannot move quickly enough, or the light source loses power. Semiconductor equipment is therefore not judged by whether it can produce one successful exposure. It is judged by whether it can repeat the same exposure across thousands of wafers with stable yields and predictable maintenance.

This distinction is central to understanding both ASML’s advantage and China’s challenge. Public diagrams can reveal a system architecture. Academic papers can explain optical principles. Engineers can disassemble components and measure dimensions. None of those actions automatically produce a tool that can operate in a fab for years. The commercially relevant product is a controlled process environment: optics, illumination, vibration isolation, wafer handling, metrology, computational correction, software and service all working together.

Lithography also becomes harder as feature dimensions shrink. A simplified expression for optical resolution is proportional to the wavelength of the light divided by numerical aperture. The semiconductor industry therefore spent decades shortening wavelength, increasing numerical aperture and improving process-control techniques. Each improvement created new problems in materials, optics, light generation and contamination. The result was not a smooth march from one machine to the next, but a series of industrial bets in which companies could disappear after choosing the wrong technical path.

ASML is therefore more accurately described as the system integrator at one of the most sensitive conversion points in the semiconductor production chain, rather than merely as a manufacturer of expensive cameras. Its machines transform optical and mechanical precision into manufacturable transistor density. In the AI era, the consequence reaches far beyond mobile processors. GPUs, CPUs, custom accelerators, advanced DRAM and HBM all depend on lithography at different process layers. The equipment sits upstream of the entire compute stack.

How Lithography Consolidated from American Pioneers to a Single EUV Supplier

This is also why China’s challenge cannot be measured only by whether a domestic scanner can reproduce a published optical design. China must rebuild the missing continuity: component suppliers, systems engineering, fab support, customer trust, production data and enough protected demand to finance repeated generations. Its domestic market can create a learning environment, just as COMAC’s home market can support aircraft development. But lithography’s much lower tolerance for instability means that the path from a usable domestic tool to a frontier-equivalent production platform is likely to be longer and more unforgiving.

The monopoly is therefore partly a result of historical selection. ASML survived enough transitions to become the repository of capabilities that had once been distributed across an industry. Its organization now contains Dutch systems integration, German optical technology and substantial American lithography and light-source engineering. A challenger is not trying to copy a purely Dutch machine. It is trying to reconstruct an international industrial coalition that the market spent decades consolidating into one company.

This created several reinforcing loops. Revenue and customer relationships from DUV helped finance EUV. Early EUV placements generated operating data. Operating data improved availability and throughput. Improved economics encouraged customers to add more EUV layers. Larger orders justified further investment by ZEISS, Cymer and hundreds of specialized suppliers. Each turn of the loop made it harder for a rival without an installed base to enter, because the rival would need production customers in order to improve the tool, while production customers would refuse the tool until it had already improved.

Most corporate moats describe an advantage over existing competitors. ASML’s EUV moat is deeper because the closest historical competitors were removed from the frontier before the technology reached economic maturity. The American pioneers had already fragmented, exited or been absorbed. Nikon and Canon retained valuable DUV and alternative-lithography capabilities, but did not carry a commercial EUV scanner through the same learning cycle. When chipmakers finally began committing billions of dollars to EUV fabs, ASML was not one of three qualified suppliers. It was the only supplier with a tool, a support organization, a supplier network and years of shared process development.

The structural consequence of an equivalent rival’s disappearance

This was the decisive strategic difference. Nikon and Canon remained important lithography companies, but neither assembled and financed an equally broad coalition through the full period between EUV research and profitable high-volume production. ASML became the place where German optics, American light sources and scanner engineering, European research, and the process knowledge of the world’s leading chip manufacturers were integrated into one roadmap.

ASML then reinforced that network through acquisitions and customer financing. SVG added American scanner, mechatronics and optical capability. Cymer added the San Diego light-source organization that became central to generating usable EUV power. ZEISS remained the indispensable optics partner. Intel, TSMC and Samsung contributed capital and process access through ASML’s customer co-investment program. Research institutes and material suppliers helped solve masks, resists, metrology and contamination problems.

The company also understood that no single corporate laboratory could industrialize EUV. In Europe, ASML formed the EUCLIDES consortium with ZEISS and other research partners. It later joined forces with the American EUV LLC program, which connected U.S. chipmakers and national-laboratory research. ASML’s own history of moving EUV from laboratory to fab shows that American, European and Asian capabilities were progressively assembled around one commercial integrator.

ASML made the opposite choice. Because lithography was its core business rather than one division inside a diversified camera and imaging group, failure to reach the next frontier threatened the company’s long-term position. That concentration created risk, but it also created commitment. ASML could not protect itself by retreating to a larger consumer franchise. It had to make the next scanner work.

Nikon continued to develop DUV and immersion platforms, where it had established engineering competence and a visible market. Canon increasingly pursued a different route through nanoimprint lithography, a method that mechanically transfers patterns rather than projecting them through an EUV optical system. Canon’s commercialization of its FPA-1200NZ2C nanoimprint platform illustrates that its strategic response was not to reproduce ASML’s EUV architecture, but to seek an alternative patterning path.

The phrase “Japan only made DUV” can be misleading if it suggests a lack of scientific understanding. Japanese companies and research programs investigated next-generation lithography, and Nikon remained technically capable in advanced ArF systems. The decisive difference was commercialization strategy. EUV was not a normal successor in which an existing lens and light source were gradually improved. It required a new wavelength, reflective rather than transmissive optics, vacuum operation, new masks and resists, a plasma light source and an entirely new contamination-control regime. The project demanded years of spending before customers could know whether a production-worthy tool would ever exist.

Why Japan remained in DUV while ASML carried EUV to production

By the 1980s and 1990s, the center of commercial lithography power had therefore moved away from the American pioneers. Nikon became the benchmark supplier across important generations, Canon remained a serious competitor, and ASML was the smaller European challenger. This is important for judging the current monopoly: ASML did not inherit an uncontested market. It had to defeat companies that already possessed stronger brands, larger installed bases and proven optics.

The Japanese advantage was not simply that camera companies knew how to polish glass. They were able to integrate lenses, stages, controls, manufacturing quality and service into tools that fabs trusted. In a production environment, a scanner that prints a slightly better test pattern but stops frequently can be less valuable than a system with stable output, predictable maintenance and fast support. Nikon and Canon helped shift competition away from laboratory performance alone and toward total manufacturing economics.

Nikon and Canon rose from a different industrial base. Both had deep experience in optics, precision manufacturing and the disciplined production of complex imaging systems. They also operated close to Japan’s rapidly expanding semiconductor manufacturers during the memory boom. Nikon shipped Japan’s first commercial stepper in 1980 and developed successive generations of i-line, KrF and ArF systems. Its official semiconductor lithography history shows the progression from early steppers to ArF immersion scanners. Canon entered semiconductor lithography in 1970 and built its own long line of aligners and steppers.

Japan won the DUV era through reliability and industrial execution

Part of the American capability survived, but not as an independent national champion. Perkin-Elmer’s Wilton lithography operations evolved through Silicon Valley Group and its Micrascan platform. ASML acquired SVG in 2001, bringing the Connecticut site, optical expertise and step-and-scan engineering into its own organization. ASML’s account of its Wilton history traces this lineage from Perkin-Elmer to SVG and then to ASML. The United States therefore did not lose every underlying capability; a substantial portion was absorbed into the transatlantic industrial system that ASML came to coordinate.

American lithography firms did not disappear because the United States forgot optical physics. They lost continuity at the company level. Some struggled to convert pioneering designs into consistently reliable volume tools. Others were divisions of broader organizations whose capital could be redirected when lithography became too cyclical or expensive. Customer trust migrated toward vendors that could combine optics with repeatable manufacturing and intensive field support. Once a supplier lost a leading-node design win, it also lost access to the production data and customer roadmap needed to build the next generation.

But the competitive requirement changed. A lithography tool stopped being primarily an optical invention and became a high-volume production system that had to run around the clock. Customers increasingly evaluated not only resolution, but availability, defectivity, overlay stability, service response, spare-parts logistics and the speed at which the vendor could correct problems across an installed fleet. The semiconductor cycle also became more punishing. Each generation required more R&D capital, while a downturn could remove the revenue needed to finish the next platform.

Perkin-Elmer’s Micralign became one of the industry’s first commercially important projection-scanning platforms. GCA’s stepper architecture then addressed the need for higher resolution by exposing a smaller field and moving the wafer from die to die. In the early period, American equipment makers were not followers. They defined the market.

Modern semiconductor photolithography emerged from American laboratories and equipment companies. Bell Labs engineers adapted photoengraving methods to semiconductor fabrication in the 1950s. Fairchild engineers used step-and-repeat cameras to replicate transistor patterns across wafers. The David W. Mann division of GCA became an early commercial supplier of step-and-repeat reduction equipment, while Perkin-Elmer developed projection systems that avoided the defect problems created when a mask physically touched a wafer. The Computer History Museum’s semiconductor timeline records how these American systems established the basic architecture of commercial optical lithography.

The United States created the commercial lithography industry

This history matters because ASML’s deepest moat is not a single mirror, patent or light source. It is the cumulative result of competitor extinction. Every major transition removed firms that could no longer finance the next machine, manufacture it reliably, support it inside fabs or persuade customers to risk a production node on an unproven platform. By the time EUV became commercially necessary, the field no longer contained several equally prepared companies. It contained one surviving systems integrator connected to the necessary suppliers, research programs and customers.

ASML’s present position can look inevitable when viewed from the end of the story. It was not. The lithography industry once contained multiple American, Japanese and European competitors, and the United States created much of the commercial foundation on which later scanners were built. Japan then converted optical and manufacturing discipline into market leadership. ASML entered from behind, survived repeated technology transitions, absorbed parts of the remaining American capability and eventually became the only supplier to commercialize projection EUV at high volume after a development cycle that lasted far longer than a normal corporate planning horizon.

How ASML Won the Immersion Transition

By the time the industry reached the 193-nanometer generation, the supplier field had already narrowed. Nikon was the established leader, Canon remained a serious optical competitor, and ASML was the challenger trying to turn its TWINSCAN architecture and customer partnerships into a durable advantage. The next technical decision—whether to continue toward 157-nanometer exposure or extend 193-nanometer lithography through immersion—would determine which companies entered the next phase with a viable platform.

The industry’s roadmap eventually approached a wall. Deep-ultraviolet systems using 248-nanometer and later 193-nanometer light enabled further scaling, but the next expected move toward 157-nanometer lithography created severe material problems. Light at that wavelength was absorbed by air and by many conventional optical materials. Calcium-fluoride optics introduced technical complications, costs rose and the path to high-volume manufacturing became uncertain.

Immersion lithography changed the roadmap by placing a thin layer of ultra-pure water between the final lens element and the wafer. Because light travels differently through water than through air, the medium increases numerical aperture and improves effective resolution without requiring an entirely new 157-nanometer light source. The concept sounds simple only after it works. In practice, a high-speed scanner had to control a moving film of water without bubbles, contamination, vibration or defects while wafers traveled through the exposure stage at production speed.

TSMC’s Burn Lin was one of the most important advocates for the immersion approach, and the partnership between toolmakers, research institutes and chip manufacturers accelerated its validation. ASML’s own history describes how early prototypes evolved into the TWINSCAN XT:1250i, how an “immersion hood” controlled the water under the lens, and how defectivity and wafer speed improved through repeated engineering. TSMC demonstrated functional 90-nanometer chips using immersion equipment in 2004. By 2006, ASML had brought immersion lithography into volume production.

The strategic lesson is not simply that ASML selected the correct wavelength. It is that ASML converted a physical insight into an industrial platform faster than rivals. It aligned a new optical architecture with customer roadmaps, process development, defect-control learning and a dual-stage machine design. Nikon’s strengths in optics did not automatically produce the same system-level transition. A lithography vendor must coordinate the whole factory interface, not only the lens.

Immersion DUV became one of the most commercially important technologies in semiconductor history. It supported advanced nodes through increasingly elaborate forms of multiple patterning and remains essential for many layers even in fabs that use EUV. ASML’s projected 2026 immersion capacity of roughly 130 systems is evidence that EUV did not make DUV obsolete. The two technologies coexist because a modern chip uses different lithography tools for different patterning tasks, and because mature and specialty semiconductors continue to require large amounts of DUV capacity.

This history also provides the standard for evaluating China’s new immersion effort. The important milestone is not that water can be placed beneath a lens. The milestone is whether the entire scanner can control defects, overlay, throughput and uptime in a real production environment. ASML’s advantage was created during the transition from concept to industrial repeatability. Public reporting indicates that China is entering that stage, but production performance has not yet been independently established.

The EUV Bet and the Construction of an Industrial Monopoly

Immersion extended 193-nanometer lithography, but it did not remove the underlying resolution problem. As leading-edge features became smaller, manufacturers relied on multiple patterning: dividing a dense design into several less-dense exposures and process steps. This kept DUV economically useful but increased the number of masks, etch cycles, cleans, measurements and alignment operations. Each additional pass added cost, cycle time and opportunities for defects.

Extreme-ultraviolet lithography attempted a much larger jump. Instead of 193-nanometer light, EUV uses a wavelength of 13.5 nanometers. That change is not an incremental upgrade to a DUV scanner. EUV light is absorbed by air and by conventional lenses, so the entire optical path must operate in vacuum and use multilayer reflective mirrors. The source is generated by firing high-powered laser pulses at fast-moving droplets of molten tin, creating plasma that emits EUV radiation. ASML says the process can occur up to 50,000 times per second.

Every part of this architecture created an industrial problem. The tin droplets had to be formed, positioned and hit consistently. Debris from the plasma had to be managed so it did not contaminate the collector. Mirrors had to achieve extraordinary surface precision and retain performance under high-energy exposure. Because each reflection loses light, the source needed enough power for economically acceptable wafer throughput. The wafer stage had to move rapidly while maintaining nanometer-scale positioning. Metrology and computational correction had to detect and compensate for distortions across the system.

ASML did not solve these problems alone. Its monopoly is the visible surface of a distributed industrial coalition. Cymer developed critical light-source capabilities and was acquired by ASML in 2013. ZEISS supplied the optical systems and developed manufacturing and measurement technologies that did not previously exist. Research organizations such as imec connected equipment development to process experiments. Intel, TSMC and Samsung participated as customers, technical partners and financial supporters. In 2012, ASML’s customer co-investment program brought capital from major chipmakers into the EUV and next-generation roadmap.

This arrangement created a form of shared dependence. Chip manufacturers needed ASML because no other vendor could provide EUV. ASML needed chip manufacturers because a new lithography generation could not be validated without their process knowledge, wafers, masks, materials and capital. Suppliers needed ASML’s volume and roadmap visibility to justify highly specialized investment. The ecosystem became difficult to duplicate because the value did not reside in one patent or component. It resided in the interfaces among organizations.

The result was an industrial monopoly rather than a conventional pricing monopoly. ASML became the only commercial supplier of EUV systems, but it could not simply maximize price without regard to customer economics. If lithography made advanced nodes uneconomic, customers would slow adoption, extend older processes or redirect capital toward packaging and architecture. ASML’s power therefore depends on maintaining a credible cost-per-wafer improvement for customers, not merely on possessing a machine no one else can make.

The EUV system became a control point for advanced logic and advanced DRAM. ASML’s NXE platform supports high-volume production at the leading nodes, while DUV systems continue to print less-critical layers. This means EUV adoption does not replace ASML’s older business so much as add a more valuable layer to it. A leading-edge fab becomes a mixed fleet of EUV, immersion DUV, dry DUV, metrology and software, creating both new-system revenue and a growing installed-base service opportunity.

China’s exclusion from EUV therefore matters at two levels. It limits access to the most efficient patterning method for certain advanced layers. It also excludes Chinese fabs from the production-data loop in which next-generation tools, materials and process techniques are jointly refined. The missing asset is not only the machine. It is participation in the frontier manufacturing network.

ASML’s Structural Moats

Optical precision

The first moat is the optical system. In DUV, lenses must transmit and shape light while controlling aberrations across the field. In EUV, conventional lenses are impossible, so multilayer mirrors guide and reduce the pattern. ZEISS has spent decades developing the fabrication and metrology required for these components. High-NA EUV pushes the challenge further: its illumination and projection optics contain tens of thousands of parts, and the measurement systems used to verify mirror quality are themselves enormous precision machines.

Optics are not a static specification. Performance must remain stable under heat, vibration, radiation and repeated operation. A mirror that is sufficiently smooth in a laboratory but deforms under thermal load does not create a commercially useful scanner. The relevant capability includes material purity, coating, polishing, assembly, sensing and active correction.

Light-source engineering

The second moat is light generation. DUV systems require powerful, stable excimer laser sources. EUV requires the conversion of laser energy into usable 13.5-nanometer radiation through tin plasma. Source power affects throughput. Stability affects dose control. Debris affects maintenance. A successful light source therefore combines plasma physics, lasers, vacuum engineering, contamination control and serviceability.

China can develop individual laser capabilities without immediately reproducing ASML’s source economics. A source must operate at high duty cycles, connect to the scanner’s control system, maintain predictable output and be serviceable within fab schedules. The challenge is not generating EUV photons once. It is generating enough controlled photons every hour for commercial production.

Motion, overlay and mechatronics

The third moat is precision motion. The wafer stage must accelerate, scan and reposition while keeping the wafer aligned with the optical image. Overlay errors compound across layers and across multiple patterning steps. Throughput requires speed; resolution requires stability. These objectives conflict. Increasing motion can increase vibration and heat, while slowing the system reduces economic productivity.

ASML’s dual-stage architecture helps separate exposure and measurement tasks, but its advantage depends on sensors, actuators, control algorithms and accumulated calibration data. A domestic challenger may reach nominal resolution while remaining far behind in wafers per hour or overlay consistency. In a fab, those gaps can matter more than the smallest feature printed during a demonstration.

Supply-chain specialization

The fourth moat is a supplier network built around components with extremely small markets and extreme requirements. The economic logic is unusual. A supplier may develop a unique module for only a limited number of systems per year. It invests because ASML provides a long roadmap, engineering support and access to the world’s most advanced customers. A new entrant must either convince the same suppliers to support a competing platform, develop domestic substitutes, or redesign the system around different components.

Export controls make this harder by targeting not only complete scanners but also specific technologies, software and support. At the same time, controls create demand for substitution. The same restriction that delays a Chinese program can create a policy-supported customer base if domestic equipment achieves acceptable performance. This dual effect is one reason the long-term outcome cannot be inferred from short-term technical gaps alone.

Customer co-development

The fifth moat is the connection to customer roadmaps. Lithography systems are developed years before a process enters volume production. Tool requirements depend on transistor architecture, mask strategy, materials, resist chemistry and process windows. ASML can prioritize engineering because it sees the roadmaps of leading logic and memory customers. Customers can influence the product because they are deeply involved in qualification.

This relationship creates information advantages. A challenger without frontier customers may not know which performance bottleneck matters most at the next node. It can optimize the wrong variable. China can partially solve this problem through domestic fabs, but those fabs do not yet represent the full frontier of global manufacturing. They provide a valuable learning loop, but not the same roadmap position as TSMC, Samsung and Intel.

Installed-base data and service

The sixth moat is the operating history of the installed fleet. Every machine produces maintenance records, error logs, calibration data and upgrade opportunities. ASML converts those observations into software changes, field options, replacement parts and next-generation designs. In 2025, net service and field-option sales reached €8.2 billion, up 26.2% from the prior year. In the second quarter of 2026 alone, Installed Base Management revenue was €2.762 billion.

This revenue is strategically important because it turns system sales into long-duration customer relationships. A scanner can remain in production for many years and receive performance upgrades over time. For China, dependence on ASML therefore extends beyond the purchase date. The machine remains connected to a foreign service, software and spare-parts system. A domestic tool can be strategically valuable even before it matches ASML’s performance if it begins localizing that operating relationship.

Competitive attrition and path dependence

The seventh and deepest moat is historical continuity. ASML is not merely ahead of several equivalent EUV competitors; it is the surviving integrator after American pioneers exited or were absorbed and Japanese rivals did not commercialize projection EUV at high volume. That difference matters because a rival cannot buy decades of uninterrupted customer access, supplier investment and production learning in a single transaction.

Each lithography generation finances and teaches the next one. DUV cash flow funded EUV development. Installed tools created field data. Customer process roadmaps identified which specifications mattered. Supplier volume justified dedicated factories and metrology systems. Once this chain was concentrated around ASML, a new entrant faced a circular barrier: it needed qualified customers to obtain the data required for qualification. This path-dependent structure is the moat beneath the optical, mechanical and service moats described above.

China’s Dependence on Imported Lithography

China entered the current technology conflict with a large and uneven semiconductor manufacturing base. It had built substantial capacity in mature and specialty nodes, expanded foundry investment, and purchased significant volumes of foreign equipment. ASML’s DUV systems were central because they could support a wide range of products: power management, image sensors, automotive chips, connectivity devices, memory and, through advanced multiple patterning, some leading-edge logic.

This purchasing wave created an apparent form of strength. Chinese fabs accumulated equipment, trained engineers and expanded output. But the same process created structural dependence. The fleet required imported parts, software, calibration, upgrades and field service. Even when a tool was physically located in China, part of its operational sovereignty remained outside the country.

ASML has not publicly reported delivering a commercial EUV system to a Chinese customer. The restriction emerged before the broader wave of semiconductor controls and became one of the clearest examples of allied technology coordination. Later Dutch measures required export authorization for specified advanced semiconductor manufacturing equipment, including certain DUV technologies, while additional controls expanded to other process categories. The policy was not formally described as a universal ban on every lithography system, but it narrowed the set of tools available to Chinese customers and increased uncertainty around future service.

This matters because a fab is not a warehouse of independent machines. Capacity depends on the full process line. A shortage of one critical tool can constrain the utilization of many others. Multiple patterning can increase demand for lithography, etch, deposition, cleaning and metrology simultaneously. If a restricted scanner becomes a bottleneck, the value of the surrounding equipment is reduced.

The resulting problem is broader than access to the latest node. China needs equipment for at least three objectives. It needs mature-node capacity for a large domestic industrial economy. It needs advanced-node experimentation for telecommunications, AI and strategic systems. It also needs the ability to maintain the installed fleet if external support becomes less reliable. A domestic immersion system can address portions of all three objectives even if it remains far from EUV.

The reported 2026 Chinese program is therefore best understood as a sovereignty project rather than only as a product launch. The first units may be less important for immediate wafer output than for creating a national integration platform. They give component suppliers a target architecture, fabs a machine to test, and public funders a measurable path from research spending to industrial deployment, while still leaving commercial performance uncertain.

How China Produced Advanced Chips Without EUV

The absence of EUV does not make advanced chip production physically impossible. It changes the cost and complexity of the process. Immersion DUV can print finer effective patterns through multiple patterning, in which a dense layer is decomposed into several masks and process sequences. Each sequence exposes a less-dense pattern, and the combined result produces smaller pitches than a single exposure could achieve.

This method supported the global semiconductor industry before EUV reached production maturity. It remains technically powerful. The limitation is economic. More masks mean more lithography passes, more deposition and etch operations, more metrology and a longer cycle time. Overlay must be controlled across every step. Defect opportunities multiply. A process may produce functional chips while delivering lower yield and lower throughput than a comparable EUV-enabled route.

SMIC’s production of advanced chips for Huawei demonstrated that DUV can be pushed further than a simplistic “no EUV, no advanced node” claim suggests. But a functional device is not the same as a globally competitive manufacturing platform. The important questions are die size, yield, volume, cost, performance consistency and the ability to repeat the process across product generations.

Multiple patterning also consumes equipment capacity. If one layer requires several lithography cycles instead of one, the same output may require more scanners or longer utilization of each scanner. This raises the strategic value of China’s existing DUV fleet and increases the incentive to build domestic replacements. A local immersion tool does not need to print the most advanced layers immediately to release capacity elsewhere in the fab.

This is one reason the domestic tool could first appear in mature or less-critical process layers. Qualification can begin where the economic cost of lower performance is manageable. Engineers can then improve overlay, defectivity and uptime before attempting more demanding patterns. The path may resemble progressive insertion rather than a direct replacement of ASML’s best immersion system.

What China’s Immersion DUV Breakthrough Actually Means

Reuters reported in July 2026 that Shanghai Aishengna Electronic Technology Group, a state-backed company, had begun production of domestically developed immersion DUV tools after incorporating teams associated with Yuliangsheng and Shanghai Micro Electronics Equipment. The reported plan called for about five systems in 2026 and approximately twenty in 2027, with expected deliveries to SMIC, Hua Hong and CXMT. The report also emphasized that the machines required further testing and remained far from matching ASML’s competing systems.

The most important word is not “mass production.” It is deployment. Five tools are too few to change global supply. They are enough to create several production-learning sites. If the machines enter real fabs, Chinese engineers can observe vibration, contamination, resist behavior, lens heating, stage drift, failure patterns and service requirements. Each failure can be translated into a design change or a supplier specification.

The program also creates a coordination mechanism. A domestic scanner requires local optical components, lasers, stages, sensors, control electronics, software, metrology and materials. Some modules may initially depend on foreign inputs, but the complete machine tells policymakers and suppliers which substitutions matter most. Without an integrating platform, domestic component programs risk becoming isolated demonstrations. With a platform, they can be measured against system performance.

The near-term commercial threat to ASML remains limited. ASML expects immersion capacity around 130 systems in 2026 and plans to increase that capacity by roughly 30% in 2027. Its tools have established throughput, overlay, uptime and customer support. A newly deployed Chinese system cannot immediately replace that industrial record. Even Chinese fabs may prefer ASML where licenses and supply remain available because yield economics can outweigh strategic preference.

The long-term meaning is different. China may now possess a platform through which state support can be converted into fab-level learning. This resembles the industrial-density mechanism described in Hi K Robot’s analysis of U.S.–China AI and robotics competition: protected deployment can create repeated cycles of installation, failure, modification and reinstallation. In robotics, a lower-performing machine may still complete simple tasks. In lithography, the tolerance is much tighter, but the learning mechanism remains relevant.

The correct conclusion is therefore conditional. China has not demonstrated an ASML-equivalent immersion platform. It may have crossed a more foundational threshold: from isolated prototypes toward a tool that can enter an industrial feedback loop. Whether that loop compounds depends on measurable results that have not yet been disclosed.

The COMAC Problem

The term “COMAC problem” is useful because it shifts attention away from a binary contest between products. China’s Commercial Aircraft Corporation did not need to build the world’s best passenger aircraft before it changed the structure of the Chinese aviation market. A domestic platform could receive orders from state-linked carriers, attract suppliers, develop certification experience, train maintenance personnel and create a political reason to reduce dependence on Boeing and Airbus.

Hi K Robot previously used this concept in its analysis of NVIDIA, AI infrastructure dependency and China’s compute-sovereignty problem. In NVIDIA’s case, the hard part is not only manufacturing an accelerator. It is reproducing the software, libraries, networking, developer tools, cloud integration and customer trust surrounding the hardware. A domestic GPU can exist without replacing the operating ecosystem.

ASML’s version is different. The barrier is not primarily a software developer ecosystem. It is an industrial operating system built from precision optics, light sources, motion control, process integration, fab data, field service and supplier specialization. The domestic machine must not only execute a design. It must survive the physical world of semiconductor production.

Yet the strategic threshold can still be lower than global competitiveness. A Chinese tool can create value by performing one of five roles:

• replacing imported tools on mature or less-critical layers;
• providing a fallback option if foreign deliveries or service are interrupted;
• creating a test platform for domestic components and materials;
• training engineers and service organizations around a local architecture;
• giving Chinese fabs leverage in negotiations over imported equipment and support.

None of these roles requires the machine to match ASML’s highest-end throughput or overlay immediately. They require it to be useful enough that repeated deployment generates more capability than it destroys in yield and cost. This is the point at which a protected market becomes an engineering asset rather than only a subsidy mechanism.

For ASML, the problem begins when China’s domestic system reduces the marginal value of dependence. Suppose Chinese fabs continue buying ASML whenever possible but assign domestic tools to selected layers. ASML may still retain the premium segment, yet its share of incremental capacity can decline. Suppose domestic suppliers begin servicing parts of the installed fleet or providing substitute components. ASML may still sell systems, yet its recurring service relationship can weaken. Suppose local tools are used mainly for training and experimentation. ASML may retain revenue, yet China’s future ability to absorb tighter controls can improve.

This is why the reported five systems matter more than their revenue equivalence. The machines can become nodes in a national learning network. The relevant comparison is not five versus 130 in one year. It is whether five become twenty, whether twenty generate reliable process data, and whether that data improves the next generation. The COMAC problem is a compounding problem.

There is also a geopolitical feedback loop. Export controls make imported tools scarcer and increase the cost of failure for domestic equipment. At the same time, they strengthen the political mandate for Chinese fabs to accept early domestic tools. A commercial buyer in an open market may reject an immature scanner. A strategic buyer under supply risk may accept lower productivity in exchange for future sovereignty. This is the mechanism that can break the chicken-and-egg barrier described earlier: a strategic buyer does not wait for the tool to prove itself under ordinary commercial standards before providing the production environment in which it can improve. This does not make the engineering easier, but it changes the adoption threshold.

The semiconductor system is therefore moving toward the two-operating-system world described in earlier K Robot Analysis. One system is organized around U.S.-allied technology, Dutch and Japanese equipment, Taiwanese and Korean manufacturing, and access to frontier process learning. The other is attempting to construct domestic substitutes under constraint. ASML remains the dominant supplier inside the first system. The COMAC problem is that the second system may become functional before it becomes equal.

Where the COMAC Analogy Breaks Down

The analogy is not a prediction that lithography will follow commercial aviation. Four differences make semiconductor equipment more difficult.

Lower tolerance for inferior performance

An aircraft with shorter range or higher fuel consumption can still serve selected routes. A lithography tool with insufficient overlay, uptime or defect control may be unusable even for an apparently less demanding layer. This does not contradict the possibility of beginning at a mature process node, where wider design rules and less aggressive patterning can provide a more forgiving qualification environment. The distinction is between entering a mature-node production flow and assuming that a supposedly less-critical layer within an advanced-node flow can tolerate a scanner that fails minimum overlay, uptime or defect-control thresholds. Performance gaps can still translate directly into lost wafers, unstable yield and uncompetitive cost.

The restricted components are close to the core

COMAC could assemble aircraft using important foreign systems while developing domestic capability over time. In lithography, many of the most difficult elements—high-end optics, light sources, control systems and metrology—are themselves strategic and potentially restricted. A Chinese scanner must either localize the hardest modules or remain vulnerable to the same external pressure it is meant to solve.

Certification comes from production economics

A government can influence airline purchasing, financing and route allocation. It cannot administratively create semiconductor yield. Fabs can be instructed to test domestic tools, but the resulting chips must still meet performance and cost requirements. Policy can create learning opportunities; it cannot repeal process physics.

The target keeps moving

Commercial aircraft generations improve, but older aircraft can remain economically useful for decades. Lithography’s frontier is tied to process scaling. As China works to industrialize immersion DUV, ASML is scaling low-NA EUV and introducing High-NA EUV. The gap can narrow in one category while widening in another.

These differences do not invalidate the COMAC framework. They define its boundary. China can use a protected domestic market to build equipment capability, but success must be measured by fab performance rather than by the existence of a national champion.

Why DUV Progress Does Not Translate Directly into EUV

Immersion DUV and EUV are both lithography technologies, but their core physical architectures are different. DUV transmits 193-nanometer light through optical materials and uses water to raise effective numerical aperture. EUV light is absorbed by air, water and conventional lenses. It must travel through vacuum and reflect from multilayer mirrors. DUV source expertise does not automatically produce a high-power tin-plasma source. DUV lens capability does not automatically produce EUV mirror capability.

This does not mean every lesson is lost. A domestic immersion program can build expertise in contamination control, stages, vibration isolation, wafer handling, software, metrology, supplier qualification and field service. Those capabilities are relevant to any advanced scanner. But they are enabling capabilities, not a direct bridge to commercial EUV.

The gap between a prototype and a production system is especially important. A prototype proves that a physical mechanism can operate. A pilot tool proves that multiple subsystems can work together under controlled conditions. A production tool must achieve throughput, availability, defectivity, overlay and service economics that customers can accept. EUV required decades of iteration across each stage.

It is therefore irresponsible to assign a precise date to China’s commercial EUV arrival based only on immersion news. An experimental EUV source or prototype could emerge years before a fab-ready system. Conversely, a state-backed program can progress faster than linear extrapolation if it concentrates talent, absorbs external knowledge and accepts high early costs. The correct method is to monitor disclosed subsystem performance and production deployment rather than to treat a calendar forecast as evidence.

ASML’s moat is also dynamic. By the time a challenger reproduces a previous generation, ASML’s installed base has accumulated more data and its suppliers have moved to new specifications. Catch-up is therefore not a race toward a fixed machine. It is a race against a moving industrial system.

The Revenue Exposure: DUV, China and Installed-Base Services

ASML’s financial structure explains why the Chinese development is simultaneously limited and relevant. The company is not an EUV-only business. Its revenue comes from new lithography systems across EUV and DUV, metrology and inspection, software, upgrades, spare parts and field service. China has historically been important because its fabs purchased large amounts of DUV equipment and built a substantial installed fleet.

In 2025, ASML reported €32.7 billion of total sales. Net system sales were €24.5 billion, while net service and field-option sales reached €8.2 billion. Service and field options grew faster than total sales because the installed base expanded, utilization increased for some customers, and more EUV field upgrades were delivered. This structure makes ASML more resilient than a company dependent only on annual tool shipments.

The service business also creates a specific China risk. A large installed fleet requires maintenance regardless of whether new export licenses are granted. If service continues, ASML can retain recurring revenue even as new-system sales decline. If service, software or spare parts become more restricted, the revenue impact can be larger than the loss of a single year’s equipment orders. Chinese fabs would also face greater pressure to develop third-party maintenance and domestic parts.

Public disclosures do not provide every geographic detail needed to calculate China’s service contribution precisely. Estimates that assign a specific gross margin or exact percentage of total company revenue to Chinese service are analytical scenarios, not company-reported facts. The direction of exposure is observable; the exact magnitude is less transparent.

The stronger counterweight is EUV. ASML’s 2026 outlook is being lifted by demand for advanced logic and memory, not by China. In the second quarter, total net sales were €9.326 billion, gross profit was €5.035 billion and net income was €2.918 billion. Installed Base Management sales were €2.762 billion. The company raised full-year sales guidance to €43–45 billion and gross-margin guidance to 54–56%.

ASML also said it was planning roughly 65 low-NA EUV systems and 130 immersion DUV systems of capacity for 2026, with plans to add about 30% to both categories in 2027 and investigate another increase for 2028. These disclosures indicate no immediate demand destruction in ASML’s global business. The China question is a longer-term issue of revenue composition and installed-base access inside an expanding market.

Export Controls, Spare Parts and the Risk of Forced Separation

Export controls have evolved from a focus on the most advanced systems toward a wider set of semiconductor manufacturing technologies. The Netherlands introduced a national authorization requirement for specified advanced equipment in 2023 and expanded the measure in 2025 to additional technologies. The policy remains license-based rather than a blanket prohibition on every tool to every destination, but Chinese customers face a progressively narrower and less predictable access path.

U.S. lawmakers have proposed H.R. 8170, the Multilateral Alignment of Technology Controls on Hardware Act, or MATCH Act. As of August 5, 2026, the bill had been introduced and ordered reported in amended form by the House Foreign Affairs Committee, but it had not become law. The introduced text sought closer alignment of allied controls on covered semiconductor manufacturing equipment and licensing requirements for servicing at covered facilities. Any final effect would depend on enacted text, allied government decisions and implementing regulations.

Three scenarios are analytically distinct.

Existing controls remain broadly stable

China continues receiving some lower-specification or licensed equipment while remaining excluded from EUV and selected advanced DUV. ASML’s China revenue declines as a share of total sales, but the company retains portions of the installed-base business. Domestic Chinese tools receive support but compete against imported systems where those remain available.

Controls expand across more immersion systems

Chinese fabs lose access to a wider range of new DUV tools. The immediate effect is a capacity and productivity constraint. The secondary effect is a larger policy-supported market for domestic scanners. Fabs that would otherwise reject immature equipment may participate in qualification because the alternative supply path is closing.

Service and parts are materially restricted

This is the most disruptive scenario. Existing tools could remain operational for a period, but maintenance risk would increase and software or performance upgrades could stop. ASML could lose recurring revenue. Chinese fabs and suppliers would likely expand independent maintenance, component substitution, parts harvesting and domestic service organizations. Short-term damage to Chinese output could coexist with faster long-term substitution.

This paradox is central. Controls can delay capability by denying high-performance tools and knowledge. They can also remove the commercial logic for continued dependence. The policy question is therefore not whether controls “work” or “fail” in a single dimension. It is how much time they create for allied technology leadership relative to how much substitution they induce inside China.

The incentives and costs are asymmetric but not one-sided. U.S. and Dutch policymakers seek to slow access to advanced manufacturing capability, yet wider controls can reduce supplier revenue, create friction over allied regulatory sovereignty and accelerate Chinese localization. Chinese fabs gain a stronger reason to test domestic tools, but they absorb lower productivity, qualification risk and potential yield losses. ASML preserves frontier control while facing less access to a large installed base. None of these actors receives a cost-free outcome.

The issue extends beyond ASML. A sovereign AI decision system ultimately depends on a physical computational base, as discussed in K Robot Analysis on the U.S.–China dual-system divide. Models, agents and institutional decision platforms remain constrained if the equipment required to manufacture processors and memory is externally controlled. Lithography is one of the places where abstract sovereignty encounters a measurable machine.

High-NA EUV and ASML’s Moving Frontier

ASML’s product roadmap does not depend on defending immersion DUV indefinitely; it extends the frontier. Conventional EUV, represented by the NXE platform, uses 13.5-nanometer light and a numerical aperture of 0.33. High-NA EUV raises numerical aperture to 0.55, improving resolution and enabling smaller features with fewer patterning steps on selected layers.

The change requires a new optical architecture. ZEISS developed larger and more complex mirrors, illumination systems and measurement equipment. ASML’s EXE platform must coordinate these optics with higher-precision stages, masks, resists, metrology and computational correction. High-NA is therefore not merely a more expensive version of the same scanner. It is a new production ecosystem.

The economic proposition is to reduce the complexity of future patterning. If low-NA EUV requires multiple exposures and additional process steps, a High-NA exposure may simplify the flow. But the scanner itself is more expensive, the exposure field is smaller in one dimension, and customers must qualify new materials and process strategies. The relevant comparison is total cost per good wafer, not system price alone.

High-NA reached an important milestone in July 2026 when ASML and Intel announced that selected Intel 18A layers had been dual-qualified on the EXE platform and used for a subset of Panther Lake production, with yields matched to the NXE route. This is stronger evidence than a research exposure because it connects the tool to a shipping product. It does not yet prove that every leading manufacturer will adopt High-NA at the same speed.

Broader customer insertion would be required to create volume economics across the High-NA ecosystem. Resist suppliers, mask companies, metrology vendors and fab operators invest more aggressively when they see a credible production base. One early customer can validate the technology; several high-volume customers would be needed to turn it into a standard platform.

For China, High-NA widens the frontier gap. For ASML, it creates execution and adoption risk. A monopoly supplier can still build a product before customers find it economically necessary. If low-NA EUV and multiple patterning remain sufficient for longer than expected, High-NA revenue can be delayed. The same technology that protects ASML from catch-up can pressure returns if customers postpone insertion.

Intel and TSMC: Two Different High-NA Strategies

Intel’s public process roadmap creates incentives for early High-NA use. After earlier manufacturing delays, its 18A and 14A strategies combine RibbonFET transistors, backside power delivery and renewed foundry ambitions. Early High-NA qualification can simplify selected critical layers and provide production experience, but it also exposes Intel to early ecosystem, utilization and execution risk.

Intel is also an integrated device manufacturer. It can qualify a new tool using its own products and internal process teams. That does not eliminate risk, but it creates a more controlled learning environment than a pure-play foundry that must protect multiple external customer schedules. The July 2026 Panther Lake milestone shows how Intel can use production as part of tool development.

TSMC has a different incentive structure. Its advantage rests on high-yield, high-volume manufacturing across a broad external customer base. If low-NA EUV can meet a node’s cost, yield and schedule requirements, TSMC may rationally delay broader High-NA insertion while ASML, Intel and the materials ecosystem resolve early production issues. That delay would reflect customer economics rather than evidence that the technology is unusable.

TSMC’s foundry model places a high cost on introducing equipment before it improves total cost per good wafer and schedule reliability. Public information as of August 5, 2026 did not establish a precise High-NA insertion schedule for TSMC. This uncertainty is material to ASML because a technically capable platform can still experience a slower revenue ramp if the largest customers do not yet see sufficient process-economics benefit.

For ASML, the strategic issue is customer timing. Intel can provide production data and validate the EXE system. A larger wave of adoption from TSMC, Samsung and advanced memory manufacturers would be needed to create the full revenue inflection. High-NA therefore represents both an option on future lithography intensity and a test of whether ASML can translate technical leadership into customer economics.

The divergence also illustrates why semiconductor equipment cannot be analyzed through product specifications alone. The same scanner has different value depending on a customer’s process maturity, business model, risk tolerance, node architecture and capital strategy. ASML controls the platform, but customers control the insertion decision.

AI, HBM and the Expansion of EUV Demand

The strongest current support for ASML is the widening of lithography demand across logic and memory. AI infrastructure requires more than GPUs. It requires host CPUs, network processors, custom accelerators, high-speed interfaces, advanced DRAM and HBM. Hi K Robot’s analysis of the CPU as an AI-infrastructure bottleneck showed why control processors and general-purpose compute remain essential around accelerators. Each category expands the demand base for advanced manufacturing.

Memory is especially important because EUV adoption is no longer confined to leading logic. Advanced DRAM uses EUV on selected layers to manage pattern complexity and improve scaling economics. HBM demand then amplifies the effect. HBM combines advanced DRAM dies with through-silicon vias, stacking and advanced packaging, but the memory dies still depend on front-end wafer processing. More HBM therefore means demand across lithography, deposition, etch, metrology and packaging equipment.

This transmission mechanism was developed in Hi K Robot’s earlier article on AI, HBM shortages and the semiconductor-equipment supercycle. The key point is that AI capital expenditure does not stop at the accelerator vendor. It moves upstream through foundries and memory producers into equipment orders, factory construction and supplier capacity.

ASML’s 2025 annual report described AI investment as a major driver of leading-edge logic, HBM and DDR5. In 2026, the company raised guidance as customers accelerated capacity plans. The benefit is structural because ASML does not need to predict which model provider or application will win. It benefits when the industry builds more advanced compute and memory capacity.

But the exposure is not risk-free. Equipment orders are lumpy, lead times are long and customers can overbuild. AI demand can remain strong while a particular year’s tool purchases slow because fabs have completed a capacity phase or are waiting for a process transition. HBM growth can shift spending toward packaging equipment rather than lithography at the margin. The number of EUV layers per memory generation matters more than a simple forecast of HBM units.

ASML’s position is therefore upstream but not detached from economics. Hyperscaler spending must ultimately convert into chip demand, fab utilization and customer confidence. The equipment cycle can amplify growth and later amplify pauses. The durability of ASML’s valuation depends on whether AI creates a sustained increase in lithography intensity rather than a single capacity surge.

Financial Analysis

ASML entered the second half of 2026 with stronger financial momentum than the Chinese DUV headlines implied. The second quarter produced €9.326 billion of total net sales, compared with €8.767 billion in the first quarter. Gross margin was 54.0%, net income was €2.918 billion and basic earnings per share were €7.59. The company sold 86 new lithography systems and five used systems during the quarter.

Installed Base Management was a major contributor. Revenue rose to €2.762 billion from €2.488 billion in the first quarter. This category includes service and field-option sales, so it captures both maintenance and productivity upgrades. The growth demonstrates why ASML’s economic model extends beyond shipments. Customers can add capacity or improve performance through upgrades without replacing the entire fleet.

Management raised 2026 guidance to €43–45 billion of sales and a 54–56% gross margin. Third-quarter guidance called for €11–12 billion of revenue and a 55–57% gross margin. The increase reflected stronger customer commitments across logic and memory and greater visibility into capacity expansion.

The most important financial tension is product mix. EUV and upgrade growth can raise average selling prices and support margins. China-related DUV contraction can reduce a historically important source of system demand. High-NA can create a new revenue layer, but its timing depends on customer adoption. Installed-base growth provides recurring support, but service exposure can become a geopolitical target.

R&D is another structural variable. ASML’s roadmap requires sustained spending to maintain the frontier, improve productivity, expand metrology and inspection, and industrialize High-NA. These expenditures function as the mechanism that keeps the product roadmap ahead of substitution. A lower R&D burden could raise near-term margin while weakening the monopoly’s future foundation.

The financial analysis therefore supports a balanced conclusion. China’s domestic DUV effort is not yet visible as a material competitive loss in ASML’s results. The company is experiencing an AI-driven expansion. The risk lies in how revenue composition evolves after the current capacity cycle and how much of the Chinese installed-base relationship remains accessible.

Valuation Framework

ASML’s valuation differs from that of an ordinary cyclical equipment supplier because the company combines monopoly characteristics, long-duration installed-base revenue and exposure to volatile semiconductor capital expenditure. A useful framework separates earnings power from the market multiple assigned to that earnings.

Four assumptions support a premium valuation:

• advanced logic and memory continue increasing lithography intensity;
• ASML maintains unique EUV and High-NA capability;
• installed-base revenue compounds as the fleet expands and upgrades become more valuable;
• China-related losses are offset by demand in allied markets and by higher-value EUV products.

Four assumptions can compress that premium:

• AI infrastructure spending slows before customer fab investment earns adequate returns;
• low-NA EUV remains sufficient for longer, delaying High-NA volume;
• service and DUV restrictions remove more China revenue than expected;
• domestic Chinese equipment becomes useful enough to reduce ASML’s long-term share of incremental capacity.

ASML’s official long-term framework has identified an opportunity for approximately €44–60 billion of annual revenue by 2030 with a gross margin of roughly 56–60%, based on market and lithography-intensity scenarios presented at its 2024 Investor Day. The raised 2026 outlook places the company closer to the lower end of that range earlier than previously expected, but it does not remove the scenario width. A large part of valuation depends on what happens after the current AI buildout.

Base case

Low-NA EUV and installed-base upgrades grow, memory adoption broadens, and High-NA enters production gradually. China declines as a share of sales but does not disappear. Revenue growth normalizes after the current acceleration while margins remain supported by product mix.

Upside case

AI demand drives sustained 2-nanometer and advanced-DRAM investment, EUV layers increase, and several major customers commit to High-NA. Upgrade revenue grows as customers seek productivity rather than only new tools. ASML reaches the upper half of its long-term revenue range with stronger recurring cash generation.

Downside case

Customer capital expenditure slows, High-NA insertion is delayed, and expanded controls reduce China systems and service revenue. Domestic Chinese tools do not need to reach EUV to pressure the DUV portion of the business. Earnings continue, but the market values ASML more like a high-quality cyclical supplier and less like a permanent monopoly compounder.

The valuation question is therefore not whether ASML is indispensable today. It is how long today’s indispensability can grow faster than customer alternatives, process substitutions and geopolitical fragmentation. Markets can assign a premium to durable structural control, but that premium remains sensitive to execution delays and changes in customer economics.

Counterfactual Analysis

If export controls had not tightened

China would probably still have pursued semiconductor equipment localization because industrial policy and supply-chain security predate the latest restrictions. The urgency, customer acceptance and capital concentration would likely have been lower. Chinese fabs could have continued purchasing proven ASML systems, reducing the economic incentive to qualify immature domestic tools. Controls did not create the desire for sovereignty, but they accelerated the conversion of that desire into procurement behavior.

If the domestic immersion tool fails commercial qualification

The program can still produce knowledge, suppliers and trained engineers. It would not materially reduce ASML dependence if throughput, overlay and uptime remain insufficient. A state-backed prototype has strategic value, but only production performance creates equipment sovereignty. The distinction prevents the article from treating every announced tool as a successful industrial platform.

If TSMC delays High-NA for several more years

ASML’s EUV monopoly would remain intact. The financial return on High-NA investment would be delayed, and low-NA productivity upgrades could become more important. Intel’s early use would provide data but might not create enough ecosystem volume. This counterfactual shows that technological leadership and commercial timing are separate variables.

If China reaches reliable DUV but not EUV

This may be the most plausible form of partial success. China could localize a significant portion of mature-node and selected advanced-patterning capacity while remaining excluded from the most efficient frontier. ASML would retain EUV dominance but lose some DUV growth, service leverage and geopolitical control over the Chinese manufacturing base. That outcome would validate the COMAC problem without implying technological parity.

Risks

Technology-adoption risk

High-NA may take longer to deliver customer economics than the technical roadmap suggests. Materials, masks, metrology and process integration must mature together. A delay would affect revenue timing and could reduce the expected return on a large R&D program.

Semiconductor-cycle risk

AI demand does not eliminate capital cycles. Customers can overorder, pause expansion or redirect spending. ASML’s backlog and long lead times provide visibility but can also make changes in demand appear with a delay.

Customer-concentration risk

A small number of logic and memory manufacturers account for frontier equipment demand. Their process decisions can change ASML’s product mix. TSMC’s High-NA timing, Intel’s foundry execution and memory manufacturers’ EUV layer counts are company-specific variables with system-wide consequences.

Geopolitical risk

Controls can reduce China revenue, provoke Chinese retaliation, complicate supplier access and increase compliance costs. Dutch and U.S. policy priorities may not remain perfectly aligned. ASML operates inside an alliance structure it does not control.

Chinese substitution risk

The first domestic immersion tools may fail, but the program cannot be evaluated solely as a single-company product cycle. State capital, policy-supported demand and integration of multiple research teams can sustain repeated attempts. The relevant risk is gradual erosion rather than immediate displacement.

Supply-chain risk

ASML depends on suppliers whose products are difficult to replace. This specialization creates a moat and a vulnerability. Capacity expansion must occur across optics, lasers, precision components and service personnel. A bottleneck at one supplier can constrain system output.

Valuation risk

A structurally exceptional company can still produce weak returns if the market price assumes rapid and uninterrupted growth. Multiple compression can occur without any loss of technological leadership if High-NA adoption, AI investment or margins fall below expectations.

Investment Monitoring Dashboard

Matrix Judgment

ASML remains one of the most structurally important companies in the physical construction of AI civilization. It controls the only commercial EUV platform, has moved High-NA into an early production environment, and sits inside a supplier-customer coalition that cannot be reproduced quickly by copying a machine design. Its 2026 financial acceleration indicates that AI, advanced logic and memory demand are strengthening the business rather than being displaced by China.

China’s immersion DUV program nevertheless represents a potentially meaningful change in the structure of the market. The reported systems are not equivalent to ASML’s mature immersion tools. Their performance is unverified and their scale is small. But they may create a domestic platform for fab deployment, supplier coordination and operating-data accumulation. That is a different threshold from global competitiveness, and it is the threshold that begins the COMAC problem.

The structural judgment is therefore neither “China has broken ASML’s monopoly” nor “five machines do not matter.” China has not reached the EUV frontier. It may be beginning to reduce the strategic value of permanent dependence below the frontier. ASML’s current roadmap emphasizes low-NA capacity, High-NA adoption, software, upgrades and a larger installed base.

The company’s long-term position can be evaluated through two conversion rates. The first is ASML’s ability to convert technical leadership into economically adopted capacity across logic and memory. The second is China’s ability to convert prototypes and policy support into reliable fab equipment. The distance between the two remains large, but both systems are now moving.

Counterfactual Compression

If not X. If China’s immersion-DUV program is structurally irrelevant until it matches ASML at the frontier, then partial domestic deployment below EUV parity would create no material change in dependence.

Then Y must simultaneously be true. Chinese fabs would have to gain no learning value from domestic tools, local suppliers would have to gain no qualification demand from installed systems, and export or service restrictions would have to create no incentive to accept lower productivity in exchange for supply continuity.

But Y contradicts observable constraints. The reported program is directed toward major domestic fabs; production tools historically improve through customer data and field service; and existing controls already differentiate access by equipment capability and licensing. These observations do not prove that China will achieve commercial success, but they rule out the claim that only frontier parity can affect dependence.

Epistemic Humility

Conclusion

ASML’s power is often described as a monopoly protected by physics. That description is directionally correct but incomplete. Physics defines the problem. Industrial organization determines who can solve it repeatedly. ASML’s real advantage combines optical science, light-source engineering, motion control, supplier specialization, customer process knowledge and decades of operating data.

China’s new immersion DUV effort does not remove those advantages. It does something more limited and potentially more durable: it creates a candidate platform around which a domestic equipment system can learn. The first tools may fail qualification. They may depend on foreign components. They may remain uneconomic. But if they enter fabs and improve through repeated deployment, the Chinese market can begin changing before Chinese technology reaches the frontier.

That is the COMAC problem for ASML. The structural effect does not begin only on the day a Chinese scanner equals an ASML scanner. It begins when China can allocate enough layers, customers, engineers and supplier investment to a domestic platform that dependence starts to decline.

ASML still controls the frontier. Over a 5–15-year horizon, the unresolved question is whether China must reach that frontier before it can reduce ASML’s control over the domestic semiconductor system.

Sources

• ASML — Q2 2026 Financial Results
• ASML — 2025 Annual Report: Financials
• ASML — From Perkin-Elmer and SVG to ASML Wilton
• ASML — Making EUV: From Lab to Fab
• ASML — High-NA EUV Reaches a High-Volume Logic Readiness Milestone
• Nikon — Semiconductor Lithography Systems History
• ZEISS — High-NA EUV Lithography and Next-Generation Optics
• Reuters — China Starts Production of Home-Grown Immersion DUV Tools
• Government of the Netherlands — Expanded Export Controls on Semiconductor Manufacturing Equipment
• U.S. Government Publishing Office — H.R. 8170, MATCH Act (Introduced Text)

Reproduction is permitted with attribution to Hi K Robot(https://www.hikrobot.com).