ASML: The Chokepoint Nobody Can Replicate

ASML sold 67 new lithography systems in the second quarter of 2025. Together with nine used systems and its service business, those shipments helped produce €7.692 billion in quarterly revenue. Orders reached €5.541 billion, including €2.3 billion for extreme-ultraviolet systems. A small number of machines can therefore shape billions of euros in spending by the world’s most advanced chipmakers.

The economics begin with a coordination problem. ASML combines a difficult light source, mirrors polished near the limits of measurement, vacuum engineering, high-speed stages, control software, and specialist suppliers. Each subsystem must perform inside one production tool, hour after hour, at yields that justify a semiconductor factory’s cost.

Lithography projects a chip pattern onto a light-sensitive silicon wafer. Smaller wavelengths allow smaller features, so advanced production moved from deep ultraviolet light toward EUV at 13.5 nanometers. That change required a new machine architecture rather than a routine improvement to an existing lens.

ASML’s description of the EUV source shows the difficulty. Molten tin droplets about 25 microns wide travel at 70 meters per second. One laser pulse flattens each droplet; another turns it into plasma that emits EUV light. The sequence repeats 50,000 times per second.

Air and most materials absorb EUV, so the light path operates in a high vacuum and uses multilayer mirrors instead of conventional lenses. ASML says the mirrors are polished to less than one atom’s thickness and positioned to nanometer precision. Meanwhile, the wafer stage checks and adjusts its position 20,000 times per second, targeting accuracy within a quarter of a nanometer. Failure in any part weakens the value of the rest.

ASML reported 5,150 suppliers in 2024. The count understates the structure because suppliers contribute specialized knowledge during development, not merely finished parts after a design is complete. The company says suppliers produce roughly 85% of the parts in its machines, with most supply-chain investment concentrated among about 200 critical partners.

ZEISS illustrates the relationship; it has supplied ASML optics since the late 1980s and helped make specialized EUV mirrors viable. TRUMPF contributes laser technology, while the 2013 acquisition of Cymer brought the EUV light source closer to ASML. The boundaries differ, yet the operating principle is consistent: important technical knowledge sits across a network that has learned to solve problems together.

A new entrant could purchase components from capable vendors; it would still lack decades of shared specifications, failure histories, manufacturing processes, and investment decisions. Suppliers also face an opportunity-cost test because building capacity for an unproven competitor may threaten a relationship with the customer already placing the largest, technically demanding orders.

A mirror with extraordinary smoothness has little commercial value if vibration, heat, contamination, or control errors ruin the exposure. ASML’s central task is integration: turning separate scientific achievements into a tool that a chipmaker can run at high volume.

That requirement explains the scale of continued spending. ASML invested €4.3 billion in research and development during 2024, when total net sales were €28.3 billion. It shipped 44 EUV systems that year alongside hundreds of DUV and metrology systems. R&D therefore supports a relatively small number of highly complex products rather than a mass-market unit curve.

Customers buy usable wafer output, not a collection of technical records. A faster light source matters when optics, stages, masks, software, and maintenance can convert the additional power into productive exposures. ASML controls the system specification and the final integration step, placing it at the point where supplier inventions become factory economics.

The financial result reflects that position: Q2 gross margin was 53.7%, and net income reached €2.290 billion. Customers therefore assign substantial value to the integrated result even though individual subsystems may have alternatives. The margin funds experiments with payoffs several product cycles away and gives suppliers a credible path from laboratory work to fab deployment.

Machines continue to change after installation as customers purchase service, parts, software, and field options that improve availability or productivity. ASML recorded €2.096 billion of Installed Base Management sales in Q2 2025, about 27% of quarterly revenue.

This business creates more than a smoother revenue stream because field engineers observe failure modes, component wear, process drift, and differences among customer fabs. Those observations guide upgrades and the next system design; a competitor working from laboratory specifications would begin without the operating record created by years of production use.

The installed base also lowers adoption risk for customers who already have trained staff, service relationships, process recipes, and factory layouts built around ASML equipment. A rival tool must outperform the incumbent enough to compensate for the transition, qualification work, and uncertainty attached to a new production platform.

Revenue from systems and the installed base funds the next round of development across ASML laboratories, supplier programs, field upgrades, and new platforms such as High NA EUV. In 2023, ASML shipped its first High NA system, increasing numerical aperture from 0.33 to 0.55 through a new optical design and faster stages.

The sequence makes catch-up difficult because a competitor must match the current product while ASML and its partners use revenue and field evidence to advance the next one. Time becomes part of the moat because each generation produces knowledge that informs the following generation.

Orders signal whether the loop remains economically relevant. Q2 2025 bookings included €2.3 billion of EUV demand, while quarterly net bookings overall reached €5.541 billion. Customers were still committing capital despite trade restrictions, cyclical semiconductor demand, and the cost of advanced fabs.

Capital can hire engineers, fund laboratories, and absorb years of losses. It cannot remove the need to qualify thousands of interacting decisions. A challenger must develop a source, optics, stages, control systems, contamination management, and manufacturing processes. At the same time, it must persuade suppliers and chipmakers to commit resources before the tool has proved itself.

The order matters. Customers hesitate without production evidence; production evidence requires installed tools; installed tools require customer confidence. Suppliers make a similar calculation before adding specialized capacity. ASML crossed those thresholds over decades, including long periods when EUV’s commercial timing remained uncertain.

Supplier concentration. A capability distributed across critical partners creates strength and dependence. Quality problems, capacity shortages, or financial stress at one supplier can delay an entire system.

Customer economics. Chipmakers adopt new lithography when fewer process steps, better yield, or denser chips justify the tool and factory costs. If the cost per useful wafer stops improving, technical leadership may not translate into demand.

Export restrictions. ASML disclosed that trade and national-security rules can limit shipments and service to particular customers. Restrictions reduce parts of the addressable market and complicate long-range capacity planning.

A durable chokepoint coordinates scarce capabilities and learns from every deployed system. Owning one exceptional component is insufficient when customers need the whole chain to work reliably. The practical test is whether each installation produces evidence, service revenue, and supplier learning that improve the next installation. If deployment adds complexity without increasing the organization’s ability to integrate it, the business owns a difficult product rather than a compounding system.

Historical cutoff: August 3, 2025. No event or financial result published after that date is used in this analysis.