The Veldhoven Paradox

Europe’s most strategic company sells almost nothing at home. That says more about Europe than about ASML.

Sooner or later, every conversation about European technological sovereignty turns to ASML. The Dutch company from Veldhoven is held up as proof that Europe still matters in the chip race. Yet a quick look at where its machines end up reveals a paradox: Europe owns the champion, but the champion has almost no customers in Europe.

The machines behind the chips

To see why, start with what ASML actually does. It does not make chips. It makes the machines that print them. A chip is built by shining light through a mask carrying the circuit design onto a thin disc of silicon, the wafer. The process is repeated layer after layer until billions of transistors fit on a few square millimetres. It works much like photography, and as in photography, the light is what matters: the shorter its wavelength, the smaller the circuits.

ASML sells two families of machines. Deep ultraviolet (DUV) systems are a mature technology. Extreme ultraviolet (EUV) systems print the most advanced chips, the ones that run artificial intelligence in data centres and on high-end smartphones, and no one else in the world can build them. The latest model costs around €350 million.

Europe makes chips, just not the most advanced ones

Europe does make chips, just not those. STMicroelectronics in Catania and Agrate (Italy), Infineon in Villach (Austria) and Dresden (Germany), NXP and Bosch are world leaders in power electronics, automotive semiconductors and sensors. These components are essential, as the car chip shortage of 2021 made painfully clear, but they rely on mature technologies. Making them requires DUV machines, often refurbished ones, bought in small numbers.

What Europe lacks are leading-edge fabs, the kind that buy EUV machines by the dozen. Those belong to TSMC, Samsung, SK Hynix, Intel and Micron, and they sit in Taiwan, South Korea, the United States and Japan. The only European site producing chips with EUV at volume, since 2023, is Intel’s plant in Leixlip, Ireland. One site is far too little to turn Europe into a customer that counts.

The numbers tell the story

ASML’s accounts tell the story. In 2025, the group reported sales of €32.7 billion. China took 33 percent of its machine sales, partly because Chinese firms stockpiled DUV equipment ahead of tighter export restrictions. South Korea took 25 percent, Taiwan 22 percent, the United States 12 percent and Japan 5 percent. Europe, the Middle East and Africa combined accounted for just 1 percent.

This is neither a commercial choice by ASML nor the result of some trade barrier. The machines simply go where wafers are processed.

The Chips Act’s missing fab

The first European Chips Act, adopted in 2023, was supposed to change course. Its stated goal was to double Europe’s share of the value of global chip production, from 10 to 20 percent by 2030.

Its flagship was Intel’s planned mega-fab in Magdeburg, backed by roughly €10 billion in German subsidies. Alongside Leixlip, it would have brought a second leading-edge fab into the heart of the Union, and with it steady demand for EUV machines.

In July 2025, Intel, deep in financial trouble, walked away, and the most credible route to advanced production on a European scale went with it. The TSMC-led plant under construction in Dresden will serve the car and industrial markets and will use DUV, not EUV.

The European Court of Auditors has called the 20 percent target unrealistic, noting that the Commission’s own forecast puts Europe’s share of the global value chain at just 11.7 percent in 2030.

Research is moving, production less so

Does that mean the first Chips Act failed? The Commission says no. It points to €52 billion in investment mobilised in three years, while the European Investment Bank has lent the sector €4.5 billion since 2021, €1.7 billion of it in 2025 alone.

The clearest gains, though, are in research. Five pilot lines are now running, including NanoIC, led by imec in Belgium, which works on chips beyond the 2 nanometre generation, and another on new materials for power electronics such as silicon carbide, coordinated by Italy’s National Research Council in Catania.

Research centres that used to compete now share equipment and expertise, helping companies take ideas from the lab to the factory. But a pilot line is not a production line. It proves that a process works and turns out prototypes, not millions of chips, and it needs only a handful of machines to do so.

ASML is only one piece of the puzzle

That 1 percent, however, tells only half the story. Behind every EUV machine stands far more than ASML: optics from Germany’s Zeiss, lasers from Germany’s Trumpf, and a dense web of Dutch, Belgian and German suppliers of precision components.

These are parts that almost no one else can make, and without them nobody, in Taiwan or in Arizona, can produce advanced chips. They are chokepoints, stages of the supply chain that everyone has to pass through.

But controlling a chokepoint is worth little without an industry next to it that turns that control into production and wealth.

Europe supplies the machines, others capture the value

That is why the paradox matters. Europe supplies the machines that Asia and America use to make advanced chips, which they then sell back to Europe.

The richest rewards, those earned by the companies that manufacture chips, design them and build AI processors, stay outside Europe. And the stakes are rising. By some estimates, AI chips will account for 70 percent of the market by the end of the decade. They are exactly the chips Europe does not make.

A more realistic Chips Act 2.0

The Chips Act 2.0, proposed by the Commission in June, seems to start from a more realistic diagnosis. No public policy can tell TSMC or Samsung where to build, or tell ASML whom to sell to. What policy can do is make investing in Europe worthwhile, by removing obstacles rather than picking winners.

On several points, the proposal moves in that direction.

Faster permits, stronger demand

The first is speed. Permitting and design for an advanced plant currently take on average seven and a half months longer in Europe than in competing countries. The new text sets a 12-month ceiling and a single permit application.

The second is demand. The proposal includes tools to speed up the uptake of new products, innovation procurement, and a fund that backs start-ups and small companies with equity and loans. It is also linked to the new data centres and AI gigafactories the Union is promoting.

The third concerns industrial clusters. The proposal creates an excellence label for regions where semiconductors are already at home, to make them more visible to investors, and it adds advanced packaging to the projects eligible for support.

Packaging, the assembly of several chips into a single component, has become decisive for AI. Italy already offers two examples of this kind of investment. In Catania, STMicroelectronics is building a €5 billion silicon carbide campus with €2 billion in public aid. In Novara, Singapore’s Silicon Box is building a €3.2 billion advanced packaging plant with €1.3 billion in aid.

Three problems remain

Three problems remain, however.

Europe’s energy disadvantage

The first is energy. According to DIGITALEUROPE, the European digital industry association, making chips in Europe costs 15 to 30 percent more than in Asia.

A fab draws power around the clock, and the price of electricity weighs on an investor’s decision as much as any subsidy. The proposal merely calls for efficient and sustainable production. What is missing are the tools that actually matter: long-term supply contracts signed directly with power producers, and grid connections delivered on predictable timelines.

Who pays?

The second is money. The Chips Act 2.0 sets out goals and instruments but does not say how much European money will fund them.

Everything hinges on negotiations over the EU’s 2028 to 2034 budget, which have only just begun and are expected to last at least 18 months. Commission officials themselves concede that funds will be scarcer than before and will have to be spent more selectively.

DIGITALEUROPE is calling for at least €20 billion in the next budget. As Silicon Saxony, the association representing the Dresden chip cluster, has put it, the Chips Act 2.0 is for now more a wish list than a genuine industrial policy.

The problem with national subsidies

The third is state aid, and it follows directly from the second. If European money does not materialise, subsidies stay national.

Each government puts its own aid on the table to attract a fab, and the winner is the country with the most fiscal room, not the one offering the best energy, skills and timelines.

Magdeburg shows the limits of this model. Germany had promised Intel around €10 billion, a sum few other member states could have matched, and it still was not enough. When Intel pulled out, the Union lost its only major leading-edge project with no fallback.

The Chips Act 2.0 says it wants to avoid these subsidy races, but it leaves untouched the rules that make them possible. For projects that concern the whole Union, the opposite approach is needed: fewer national subsidies competing against each other, more common resources, and in both cases transparent amounts tied to verifiable results, such as production starting by a firm date.

Europe’s strategic paradox

So for now the paradox stands. Europe’s most strategic company is European in its engineering, its suppliers and its headquarters, but global, and above all Asian, in its markets.

Whether that is a strength or a weakness will depend less on ASML than on whether Europe can once again become a place where it pays to make chips, and therefore to buy its machines.