Europe Builds Expensive Fab; Customers Ghost

Europe built the fabs. Nobody told the customers to show up.

That's the honest summary of where Europe's semiconductor strategy stands in early 2026. The EU Chips Act — Europe's attempt to reclaim a sliver of a supply chain it almost entirely handed to Asia over four decades — has catalyzed more than €80 billion in manufacturing investments, according to the European Commission. Intel was going to put a €30 billion megafab in Magdeburg, Germany. STMicroelectronics and GlobalFoundries were building a €7.5 billion joint facility in Crolles, France. Germany's federal government pledged €9.9 billion in subsidies to make Intel's project economics work. And in July 2025, Intel's new CEO Lip-Bu Tan confirmed the whole thing was cancelled — citing insufficient customer commitments, the same reason GlobalFoundries quietly slowed its participation in the Crolles project earlier that year.

The change is not cyclical. It is structural. That is how Jens Drews, GlobalFoundries' EMEA director, put it in an EE Times interview that landed with more force than the usual analyst hedging. He meant the reorientation of global semiconductor trade away from pure efficiency and toward resilience, toward allied-sourcing, toward strategic autonomy. But the phrase applies just as accurately to what went wrong on the demand side.

The EU Chips Act has mobilized more than €43 billion in public and private investment, with the Commission's own figures showing more than €80 billion in total manufacturing commitments triggered. The European Court of Auditors reviewed the program and concluded that the bloc is very unlikely to meet its stated target of 20 percent of global semiconductor market share by 2030. The Commission's own forecast projected the EU's share of global chip value growing from 9.8 percent in 2022 to only 11.7 percent by 2030. The auditors called the 20 percent target deeply disconnected from reality — and noted that reaching it would require roughly quadrupling European production capacity.

Europe's actual semiconductor strengths are narrower and more specific than the policy implies. Infineon, NXP Semiconductors, and STMicroelectronics together account for nearly half the global automotive semiconductor market, according to the Centre for European Policy Studies. These are companies that design and manufacture the chips that go into cars, industrial equipment, and power infrastructure — not the cutting-edge logic that AI training clusters require. The distinction matters because it defines both what European fabs can realistically produce and who their natural customers are.

Dresden is the case study in both the strategy's best outcome and its hard ceiling. The ESMC fab — a joint venture among TSMC, Bosch, Infineon, and NXP — is investing more than €10 billion in a facility that will produce wafers at the 28nm, 22nm, 16nm, and 12nm nodes. Earthwork was completed in spring 2025, foundation slabs laid in summer, and the first building levels were going up by autumn. Structural construction finished in late 2025, with the office building topping out on December 4. Equipment move-in is targeted for the second half of 2026, and production is expected to start in 2027. The facility is designed to process 40,000 300mm wafers per month at peak capacity and will create roughly 2,000 direct jobs plus thousands more in the surrounding supplier base. Within one zip code in Dresden, you can now find three operational or forthcoming foundries — GlobalFoundries, X-Fab, and ESMC — plus IDMs including Bosch and Infineon, and the Fraunhofer Center for Nanoelectronic Technologies. Nothing comparable exists elsewhere in Europe.

It is a genuinely impressive industrial cluster. And it will not make leading-edge chips. The ESMC site was not designed to accommodate ASML's extreme ultraviolet lithography machines — the large EUV tools simply would not fit into the structures as planned. EUV is required below 7nm. Europe has no path to leading-edge logic in this generation of fab planning. ASML, Europe's most critical semiconductor equipment company, holds roughly 90 percent of the advanced lithography market and ships its EUV machines exclusively to foundries in Taiwan, South Korea, and the United States.

The irony is sharper when you consider what capacity Europe is actually building. TSMC's Dresden JV will produce nodes that carmakers, industrial companies, and medical device manufacturers need — exactly the customers the EU Chips Act was supposed to serve. But those customers are not automatically going to shift their purchasing to European suppliers just because European capacity exists. They have existing supply contracts. They have qualification cycles. They have cost pressures. And they have a China question hanging over every purchasing decision right now.

China is expected to account for 22 percent of global semiconductor manufacturing capacity by 2030, potentially overtaking Taiwan as the world's largest chip producer. The surge is concentrated in mature nodes — exactly the 28nm and above range where European capacity is being built. We are already seeing signs of potential overcapacity in some mature-node segments, Antonia Hmaidi, a senior fellow at the Mercator Institute for China Studies, told EE Times. If Chinese producers significantly increase output, European companies could face real pricing pressure. The EU spent years and billions building the supply. China may spend less time and arrive with more volume.

There is also a structural problem in how the policy itself was designed. All of these industrial subsidies are justified as R&D subsidies, an EU policy official told EE Times. But in practice, many projects are fundamentally about expanding manufacturing capacity. The distinction matters because the approval and oversight mechanisms were built for research programs — grant applications, milestone tracking, deliverables — not for multi-billion-euro industrial commitments that can stall when an anchor customer walks. Once funds are allocated under the current rules, they cannot easily be redirected without new approvals; when a project stalls, money sits idle. Germany found out the hard way after Intel's cancellation: the federal government had earmarked €15 billion for chip industry support through 2028, with €9.9 billion pledged to Intel's Magdeburg fab. When the project collapsed, Berlin shifted €3 billion of the remaining allocation to road repairs — and the reallocation itself required restarting the approval process, because the money was locked inside a policy instrument designed for something else.

The EU's financial firepower also looks modest in global context. The €43 billion the EU Chips Act mobilizes sounds substantial until you compare it with what other regions are deploying. The U.S. CHIPS Act committed $52.7 billion plus a 25 percent investment tax credit — increased to 35 percent for fabs that break ground before 2026 — and centralized program administration in a single negotiating office, a structure that proved far more streamlined than Europe's fragmented approach. Japan committed up to ¥10 trillion, roughly €58 billion, in government subsidies and incentives by 2030. South Korea's K-Semiconductor Strategy calls for ₩622 trillion, approximately €440 billion, through 2047. China has invested more than $100 billion in semiconductors since 2014 through its state-backed Big Fund, with a new ¥344 billion tranche launched in 2024.

The EU's own contribution is further constrained by structure. The European Commission directly manages only about 10 percent of the €43 billion in public funding under the Chips Act, with the rest scattered across national programs and private investment. The Commission is responsible for only 5 percent — €4.5 billion — of the estimated €86 billion in total Chips Act funding through 2030, with the remainder dependent on member-state budgets and industry. This is not a fiscal constraint for the EU as a whole. It is a coordination and commitment problem across 27 national governments, each with their own industrial priorities.

What nobody in Brussels seems eager to discuss publicly is the demand side of the equation. In Europe, we have been very good at addressing supply, Jens Drews said. I think Europe also needs to take care of demand for capacities in Europe. That is a remarkably direct way to say that policy has been solving the wrong half of the problem.

Several first-of-a-kind Chips Act projects remain in various stages of uncertainty — waiting on customer commitments that have not materialized, or renegotiating scope after a partner pulled out. The fabs will eventually get built. Whether they have buyers when they open is a question the EU Chips Act was never designed to answer.

The global semiconductor market is expected to reach $975 billion in annual sales in 2026, per EE Times reporting. Europe will take a small and slowly growing percentage of that. The investment is real. The strategic logic is real. The gap between what was promised and what the market will bear is also real — and at this point, it would take a rather large reversal of established economics to close it.