Driven by the global race for artificial intelligence, ASML is raising its forecasts and filling its order books. But Europe accounted for none of its net system sales in both the first and second quarters of 2026, unlike in 2024, when this share was still 5%. Despite all the speeches devoted to the subject, the fact remains that organized demand is lacking, even though the continent lacks neither factories nor know-how.
The contrast is all the more striking as ASML is not going through a bad patch. The manufacturer of lithography machines, the only global supplier of EUV systems used to engrave the most advanced chips, in July raised its annual turnover forecast to between 43 and 45 billion euros. If its customers are accelerating their investments in the computing and memory capacities necessary for AI, they are not doing so in Europe.oi.
A zero that must be read before commenting on it
That being said, my figure must be handled with caution. It does not mean that ASML is no longer collecting money in Europe, nor that the continent’s factories have been emptied of their equipment. Quarterly tables measure net sales of newly recognized systems in revenue, after delivery and acceptance by the customer. Therefore, an already installed machine continues to generate maintenance, upgrades and services.
Thus, ASML’s half-year results show 407 million euros in sales in the EMEA zone, which includes Europe, the Middle East and Africa. The group does not publish net income by region. De facto, to say that it gains “nothing” in Europe, or that Europe does not represent any of its profits, would therefore be false.
But two consecutive quarters without recorded system sales, after a continued decline in the European share, remains an enlightening signal on the geography of new projects.
ASML has the orders, Europe does not place them
The reasons for ASML’s prosperity are known, leading foundries and memory manufacturers are adding capacity for the most advanced nodes. Chips for data centers, AI accelerators and ultra-high bandwidth memory require more lithography steps and, for the most sophisticated generations, EUV systems from the Veldhoven group.
This wave of investment has a very concentrated geography. It is being played out in Taiwan, South Korea, the United States and, in another technological category, China. Behind these manufacturers are the major sponsors of computing: American hyperscalers, processor and accelerator designers, smartphone players, platforms and companies that can guarantee volumes over several years.
Europe has ASML, Zeiss, imec, material suppliers, engineering skills and several producers of critical components. It holds an essential part of the global production system, but does not concentrate the customers who set the roadmaps for cutting-edge foundries.
Europe invests where it is strong, not where AI buys
It would be too easy to conclude that Europe has given up manufacturing semiconductors. Dresden tells precisely the opposite. Infineon opened its Smart Power Fab there in July, dedicated to power components, analog and mixed circuits. These chips are essential for electric vehicles, networks, industrial equipment and, increasingly, the power supply of data centers.
Also in Saxony, ESMC, the joint venture of TSMC, Bosch, Infineon and NXP, is building a factory which will produce chips on 28/22 nanometer and 16/12 nanometer processes for automobiles and industry. GlobalFoundries is also expanding its capabilities in Dresden. These projects address very real dependencies: Europe needs power, connectivity, security and control components, not just GPUs.
Their limit is elsewhere. This specialization makes the possibility of Europe becoming a third world power in semiconductors more uncertain. These projects are not organized around the demand that drives large foundries today: advanced computing and memory for AI. The chips that are used to control a motor, secure an electrical network or manage a radar are not secondary technologies. They simply don’t call for the same EUV intensity, investment levels, or machine orders as the most advanced compute nodes.
Intel provides a useful counterpoint. Ireland already has production using EUV. But an industrial exception, however important it may be, does not alone create a European market for advanced systems. Nor does it transform Europe into the center of gravity of global demand.
The customer that the Chips Act forgot
This is where the debate on subsidies finds its limit, a fab does not fill up because a State finances its building or reduces its electricity bill, but when a chip designer has a product to industrialize, when a client has qualified it and when a principal agrees to secure volumes.
The first Chips Act responded to a real emergency, namely attracting and protecting production capacities. But he first reasoned with the offer. When Europe has supported sites, production lines and research ecosystems, it has less well organized the link between its major buyers (automobile, energy, defense, telecommunications, industry, cloud) and the manufacturers who could produce the components they will need in five or ten years.
But these buyers exist, Airbus, Thales, Schneider Electric, Siemens, automobile manufacturers, telecom operators and large energy companies buy chips in considerable volumes. But they do it in global chains where the technology, availability and cost are defined elsewhere. They do not yet constitute, together, a European demand capable of giving multi-year visibility to a foundry.
Frank Heemskerk, ASML’s head of public affairs, summed up the unease during a panel discussion in the Netherlands: “We sell absolutely nothing in Europe. » His message is not just aimed at chipmakers. He asks that the customers of these manufacturers come closer to European capacities, so that industrial projects finally have an economic reason to deploy.
ASML is obviously not a neutral observer. The group has every interest in Europe installing more machines, especially the most advanced. But his observation overlaps with a broader weakness: the continent has often conceived of sovereignty as a capacity to produce, without always organizing the market which makes this capacity sustainable.
Finance the customer, not just the factory
The answer does not consist of imposing a “buy European” type slogan. Semiconductor value chains are global, and a blanket preference would be both costly and inefficient. It consists of choosing segments in which a European capacity provides clear security, resilience or performance value, then giving this capacity credible outlets.
For defense, space, electrical networks, telecommunications or certain cloud infrastructures, public procurement can integrate security of supply, traceability and continuity of service criteria. Manufacturers can engage earlier in the co-design of components with European manufacturers. Innovation aid can finance not only a production line, but also the transition from prototype to qualification, from design to volume.
It is also necessary to address the two often neglected links, design and packaging. A chip engraved in Europe but designed elsewhere, then sent outside the continent for assembly and testing, creates only partial sovereignty. The question of data centers and AI Gigafactories is part of the same logic, they are not only consumers of electricity and GPUs, but potential buyers and prescribers of the next generation of silicon.
The European Commission appears to have identified this change. His proposal for Chips Act 2.0 now emphasizes design, demand support measures and innovative public purchasing. This is a necessary development. It remains to transform this intention into contracts, specifications and order schedules. An industrial policy begins when someone agrees to buy.
The machine is European, the market is not yet
ASML’s results do not demonstrate that Europe would be absent from semiconductors. They show a specialized Europe, essential even, but still too little present where the large volumes of tomorrow are decided. The laboratories exist, the engineers are there, the factories are being built. The difficulties start after: designing the chips, aligning buyers, guaranteeing volumes and retaining value all the way to packaging and the cloud.
The real issue is therefore not whether Europe should buy more ASML machines. It already manufactures the most strategic ones in the world. It is to be understood whether it will be able to create enough European customers capable of telling manufacturers which chips to produce, for what uses and in what quantities.
