TSMC in Dresden is not the same as European chip sovereignty

The difference between hosting foreign production and owning the capability. What the European Chips Act actually buys, and what it does not.

By VastBlue Editorial · 2026-03-26 · 22 min read

Series: Reindustrialising Europe · Episode 4

The Fab That Europe Celebrated

On August 8, 2023, the Taiwan Semiconductor Manufacturing Company announced it would build a semiconductor fabrication plant in Dresden, Germany. The factory — operated through a joint venture called European Semiconductor Manufacturing Company, or ESMC — would produce chips on TSMC's 12-nanometre, 16-nanometre, and 28-nanometre process nodes. Bosch, Infineon, and NXP, three of Europe's largest chipmakers, would each take a ten per cent stake. The German federal government committed five billion euros in direct subsidies. The total investment was projected at approximately ten billion euros. TSMC would hold seventy per cent of the venture and supply the process technology, equipment specifications, and operational know-how. European politicians treated the announcement as a triumph.

Ursula von der Leyen called it a milestone for European semiconductor strategy. Olaf Scholz described it as a signal that Germany remained a globally competitive industrial location. The European Commission pointed to it as validation of the European Chips Act, the forty-three-billion-euro legislative package passed in September 2023 with the explicit goal of doubling Europe's share of global semiconductor production from approximately nine per cent to twenty per cent by 2030. Press releases multiplied. Photo opportunities were arranged. The narrative was set: Europe was getting serious about chips.

The narrative was also, in several important respects, misleading. What Europe was getting in Dresden was not semiconductor sovereignty. It was a branch factory of a Taiwanese company, producing chips designed elsewhere, using process technology developed in Hsinchu, running on equipment designed in the Netherlands and Japan and the United States, staffed — at least in its senior technical positions — by engineers trained in TSMC's own institutional system. This is not a criticism of the Dresden project. It is a description of it. And the distinction between what was announced and what was implied matters enormously for anyone trying to understand where Europe actually stands in the global semiconductor landscape.

What a Fab Actually Is — and What It Is Not

A semiconductor fabrication plant is among the most complex manufacturing facilities ever constructed by human beings. A modern leading-edge fab costs between fifteen and twenty-five billion dollars to build and equip. It contains roughly a thousand distinct process steps, performed in sequence on silicon wafers over a period of two to three months. The environment inside the cleanroom is controlled to a degree that makes pharmaceutical manufacturing look casual: particle counts are maintained below ten particles per cubic metre of air, temperature is regulated to within a fraction of a degree, and vibration is damped to levels that would be undetectable by human senses but could blur the features being printed on a chip at the atomic scale.

But a fab is, fundamentally, a machine for executing a recipe. The recipe — the process of record, in industry terminology — specifies every parameter of every step: the gas flows and pressures for chemical vapour deposition, the etch chemistries and durations for patterning, the implant energies and doses for doping, the thermal profiles for annealing, the chemical-mechanical polishing rates for planarisation. The process of record is the intellectual property. The fab is the physical infrastructure that runs it. You can own a fab without owning the process of record, just as you can own a printing press without having written the book.

This is precisely the situation in Dresden. TSMC will supply the process of record. ESMC will execute it. The European partners — Bosch, Infineon, NXP — will benefit from having a local source of fabricated wafers on mature process nodes that they currently source from foundries in Taiwan, South Korea, and increasingly mainland China. The chips produced in Dresden will go into automotive controllers, industrial sensors, power management circuits, and communications modules — the workhorses of European industry, not the headline-grabbing processors that power smartphones and AI servers. This is useful. It addresses a real supply chain vulnerability that the 2020-2022 chip shortage exposed with painful clarity, when European automakers lost billions in revenue because they could not obtain twenty-dollar microcontrollers.

But useful is not sovereign. If a geopolitical crisis were to disrupt TSMC's ability or willingness to support the Dresden fab — through export controls, a Taiwan contingency, or a shift in Taiwanese strategic priorities — the factory would not be able to develop new process technology independently. It would be a printing press without an author. The equipment would still be there. The cleanroom would still be clean. But the capability to design, develop, and iterate the manufacturing process — the actual source of competitive advantage in semiconductors — would remain in Hsinchu, 9,000 kilometres away.

The Architecture of Dependency

To understand why the distinction between hosting and owning matters, consider the supply chain that makes a modern semiconductor fab operational. It is not merely complex. It is the most geographically concentrated and interdependent industrial supply chain in human history.

The lithography tools — the machines that print circuit patterns onto silicon wafers — come almost exclusively from ASML, headquartered in Veldhoven, the Netherlands. For leading-edge production, ASML's extreme ultraviolet lithography systems are the only option. Each EUV machine costs approximately 350 million euros, weighs roughly 180 tonnes, contains over 100,000 components, and requires a light source that produces plasma at a temperature of approximately 220,000 degrees Celsius — forty times the temperature of the surface of the sun — to generate the 13.5-nanometre wavelength light needed to print features smaller than a virus. ASML is a European company, which is sometimes cited as evidence of European semiconductor strength. It is, but in a narrow sense: ASML makes the tools. It does not run the fabs that use them.

The deposition, etch, and metrology tools come primarily from American companies — Applied Materials, Lam Research, KLA — and from Tokyo Electron in Japan. The silicon wafers themselves come from a handful of suppliers: Shin-Etsu and SUMCO in Japan, Siltronic in Germany, SK Siltron in South Korea, and GlobalWafers in Taiwan. The specialty chemicals, photoresists, and electronic gases come from Japanese firms like JSR, TOK, and Shin-Etsu Chemical, and from German firms like Merck KGaA and BASF. The electronic design automation software — the tools used to design the chips that the fabs produce — comes almost entirely from three American companies: Synopsys, Cadence, and Siemens EDA (formerly Mentor Graphics, now part of a German parent but with its engineering centre in the United States).

No single country controls this entire chain. But the United States, through its dominance of equipment, EDA software, and its leverage over allies through export control regimes, exercises a degree of influence over the global semiconductor supply chain that is without parallel in any other industry. The 2022 CHIPS Act, the October 2022 export controls on advanced semiconductor equipment to China, and the subsequent coordination with the Netherlands and Japan to restrict Chinese access to EUV and advanced DUV lithography tools demonstrated this influence with unmistakable clarity. When the United States decided that China should not have access to certain chipmaking capabilities, it made that decision effective — not because it manufactures the tools, but because it controls enough chokepoints in the supply chain to enforce compliance.

Europe's position in this architecture is paradoxical. It hosts ASML, arguably the single most strategically important company in the entire semiconductor supply chain. It hosts leading producers of specialty chemicals, wafer substrates, and automotive-grade chip design. But it manufactures less than nine per cent of the world's chips, designs fewer of the leading-edge processors that command the highest margins, and operates no fab at the cutting edge of process technology — the 3-nanometre and 2-nanometre nodes where TSMC and Samsung are competing for supremacy. The Dresden fab, with its 12-to-28-nanometre process nodes, is approximately three to four technology generations behind the leading edge. This is not an accident. It reflects a strategic choice about where European resources can most usefully be deployed. But it is a choice that concedes the technological frontier to others.

What the European Chips Act Actually Bought

The European Chips Act, adopted in September 2023, committed forty-three billion euros in public and private investment to semiconductor manufacturing, design, and research across the European Union. The headline figure was arresting: forty-three billion euros is more than the annual GDP of several EU member states. The stated objective — doubling Europe's global production share from nine per cent to twenty per cent by 2030 — was even more arresting. When the Act was announced, the global semiconductor industry was investing approximately 150 billion dollars per year in new capacity. Doubling Europe's share would require not merely keeping pace with global investment but outpacing it, in a competitive environment where TSMC, Samsung, Intel, and the Chinese government were all simultaneously pouring hundreds of billions into new fabs.

The Act's three pillars revealed a more nuanced — and more modest — strategy than the headline suggested. Pillar One allocated approximately 3.3 billion euros to the Chips Joint Undertaking, a public-private partnership focused on semiconductor research and development, particularly in areas where Europe has existing strengths: automotive chips, power semiconductors, sensors, and emerging technologies like photonic integrated circuits and quantum chips. Pillar Two established a framework for recognising and subsidising "first-of-a-kind" semiconductor facilities — the category under which the Dresden TSMC fab, Intel's planned Magdeburg facility, and other projects would qualify for state aid that would normally be prohibited under EU competition law. Pillar Three created a monitoring and crisis-response mechanism to detect and respond to future supply chain disruptions.

The most consequential element of the Act was Pillar Two, because it effectively rewrite the European Union's state aid rules for semiconductors. Under normal circumstances, EU competition law strictly limits the subsidies that member states can offer to attract private investment — a principle designed to prevent a destructive race to the bottom in which countries outbid each other to lure factories with public money. The Chips Act created an exception: semiconductor fabs that qualified as "Integrated Production Facilities" or "Open EU Foundries" could receive subsidies of up to one hundred per cent of the "funding gap" — the difference between the cost of building the fab in Europe and the cost of building it in an alternative location. In practice, this meant that European governments could cover forty to fifty per cent of the total cost of a new fab, provided the project met the Act's criteria for innovation and strategic value.

This framework attracted three major commitments in its first year. TSMC's Dresden fab. Intel's announcement of two new fabs in Magdeburg, Germany, at a projected cost of thirty billion euros with approximately ten billion in German federal subsidies. And STMicroelectronics-GlobalFoundries' planned fab in Crolles, France, with 2.9 billion euros in French state support. Collectively, these projects represented the largest wave of semiconductor investment in European history.

Then the economic reality intervened. In September 2024, Intel announced that it was pausing construction of its Magdeburg fabs as part of a global restructuring driven by the company's deteriorating financial position. Intel CEO Pat Gelsinger, who had personally championed the European expansion, was under intense pressure from a board that questioned whether the company could simultaneously invest in leading-edge process technology, build new fabs in the United States under the US CHIPS Act, and expand into Europe — all while its core business was losing market share to AMD and its foundry services division was hemorrhaging money. The Magdeburg pause was not a cancellation — Intel maintained that the project was delayed, not dead — but it removed the largest single investment from the European Chips Act's portfolio and raised pointed questions about the durability of subsidy-driven industrial strategy.

Dresden and the Sovereignty Illusion

Return now to Dresden. The ESMC fab is expected to begin production in 2027 or 2028, producing chips on 12-to-28-nanometre nodes primarily for the automotive and industrial markets. At full capacity, it will process approximately 40,000 wafer starts per month — a significant volume, but modest by global standards. TSMC's largest fab in Taiwan, Fab 18, processes over 100,000 wafer starts per month on far more advanced technology. The Dresden fab will be useful. It will provide European automakers and industrial equipment manufacturers with a local source of chips that they currently import from Asia. It will create approximately two thousand direct jobs, many of them highly skilled. It will generate supply chain activity in the surrounding region. It will reduce, at the margin, Europe's vulnerability to the specific type of supply disruption that paralysed the automotive industry in 2021.

But sovereignty requires more than a factory. It requires the ability to develop the next generation of process technology. To design the transistor architectures that will define the chips of 2030 and 2035. To train the process engineers who can push manufacturing yields from ninety per cent to ninety-eight per cent on a new node. To build the ecosystem of tool vendors, materials suppliers, packaging specialists, and design houses that collectively constitute a self-sustaining semiconductor industry. None of these capabilities are created by the Dresden fab. They remain, overwhelmingly, in Taiwan, the United States, South Korea, and Japan.

Consider the human capital dimension. TSMC's competitive advantage rests not primarily on its equipment — its competitors buy similar machines from the same vendors — but on its process engineering culture. TSMC employs approximately 73,000 people, of whom roughly 12,000 hold PhD or master's degrees in semiconductor-related disciplines. These engineers have spent decades refining the manufacturing processes that allow TSMC to achieve yields above ninety-five per cent on its most advanced nodes — yields that Intel, with decades of its own experience, has struggled to match. This knowledge is deeply institutional. It lives in the calibration procedures, the defect classification databases, the equipment qualification protocols, and the thousands of small process adjustments that experienced engineers make based on pattern recognition developed over years of working with a specific tool set on a specific process flow.

A semiconductor fab without its own process development capability is a toll manufacturer, not a technology company. It can produce what others have designed but cannot create what the market will demand next. Europe is paying billions for the former and calling it the latter.

Editorial analysis

The ESMC fab in Dresden will receive its process of record from TSMC headquarters in Hsinchu. When a process issue arises — as it inevitably will, because semiconductor manufacturing involves troubleshooting issues on a daily basis — the resolution path will run through TSMC's engineering organisation, not through a European research institution. The fab's technology roadmap will be determined in Taiwan, based on TSMC's global strategic priorities, not on European industrial needs. If TSMC decides that 12-nanometre technology is no longer worth supporting because its global customers have migrated to more advanced nodes, the Dresden fab will face the same obsolescence risk as any other branch operation of a foreign parent.

Where Europe Actually Leads — and Where It Does Not

The European semiconductor industry is not weak. It is lopsided. In certain segments, European companies are global leaders. Infineon is the world's largest manufacturer of automotive semiconductors and power electronics. NXP dominates the automotive networking and security chip markets. STMicroelectronics has strong positions in sensors, microcontrollers, and silicon carbide power devices. These companies design and manufacture chips that go into virtually every car, industrial robot, and power conversion system on earth. They are formidable businesses with deep technical moats.

ASML, as discussed, occupies what may be the most powerful single-company position in any global supply chain. Without ASML's lithography tools, no one on earth can manufacture chips at the leading edge. This is not a contingent advantage — it is a structural monopoly built on three decades of research, approximately twenty billion euros in cumulative R&D investment, and an ecosystem of over 800 suppliers, many of them European, that produce the optical systems, laser sources, and precision mechanics that make EUV lithography possible. Zeiss, the German optics company, produces the mirror systems for EUV tools — mirrors polished to a flatness of less than one-tenth of a nanometre, a surface quality so extreme that if the mirror were the size of Germany, the tallest bump would be less than one millimetre high.

• ASML (Netherlands): Sole global supplier of EUV lithography — the defining tool for leading-edge chip production
• Infineon (Germany): World's #1 in automotive semiconductors and power electronics
• NXP (Netherlands): Global leader in automotive networking, security chips, and NFC
• STMicroelectronics (Switzerland/France/Italy): Leading positions in sensors, microcontrollers, SiC power devices
• Zeiss (Germany): Sole supplier of EUV optical systems to ASML — the optics that make sub-7nm possible
• Siltronic (Germany): One of five global suppliers of 300mm silicon wafers
• Merck KGaA (Germany): Major supplier of specialty chemicals and electronic materials for chip fabrication
• imec (Belgium): World's leading semiconductor R&D centre, partnering with every major chipmaker on next-generation process development

And then there is imec. Based in Leuven, Belgium, the Interuniversity Microelectronics Centre is arguably the most important semiconductor research institution in the world. Every major chipmaker — TSMC, Samsung, Intel, and increasingly Chinese firms — sends engineers to imec to work on next-generation process development. imec's pilot line runs processes two to three technology generations ahead of what is in commercial production. It is the place where the fundamental research questions of how to keep shrinking transistors get answered before the answers are transferred to commercial fabs. Europe hosts the world's premier semiconductor research centre. It simply does not translate that research into commercial manufacturing at scale.

This is the core asymmetry. Europe has world-class capabilities in semiconductor equipment, materials, research, and certain categories of chip design. What it lacks is the large-scale, high-volume advanced manufacturing capability that converts those upstream advantages into downstream products. It is as if Europe had invented the internal combustion engine, built the world's best machine tools, and trained the finest automotive engineers — but outsourced the actual assembly of cars to factories in other continents. The upstream brilliance is real. The downstream absence is also real. And the Dresden fab does not bridge this gap. It fills a specific supply chain need at mature nodes while the strategic question of leading-edge capability remains unanswered.

The Real Strategic Question

What would genuine European semiconductor sovereignty actually require? The honest answer is uncomfortable. It would require Europe to build not just fabs but an integrated ecosystem: advanced process development capability, leading-edge chip design capacity, a workforce pipeline producing thousands of semiconductor engineers per year, and a domestic customer base large enough to sustain commercially viable production volumes. It would require decades of sustained investment at levels that make the European Chips Act look like a down payment. And it would require a degree of strategic coordination across EU member states that has no precedent in European industrial history.

The Chips Joint Undertaking — Pillar One of the Chips Act — is the closest thing Europe has to a strategic investment in indigenous capability. Its 3.3-billion-euro budget funds research into areas where Europe could plausibly develop competitive advantages: photonic integrated circuits, where imec and Eindhoven University of Technology are global leaders; compound semiconductors like gallium nitride and silicon carbide, where Infineon and STMicroelectronics have strong positions; and heterogeneous integration and advanced packaging, the domain that increasingly determines system performance as transistor scaling slows. These are real technologies with real commercial potential. But 3.3 billion euros is roughly what TSMC spends on R&D every six months.

There is an alternative strategic framework that deserves consideration, even if it is less politically appealing than the sovereignty narrative. Rather than attempting to replicate the full semiconductor value chain domestically — an objective that would require investment in the hundreds of billions and a timeline measured in decades — Europe could focus on deepening and protecting its existing chokepoint positions. ASML's monopoly on EUV lithography. imec's unmatched research capability. The specialty materials and chemical expertise concentrated in German, French, and Belgian firms. The power semiconductor and automotive chip design capabilities of Infineon, NXP, and STMicroelectronics. These positions give Europe genuine leverage in the global semiconductor ecosystem — leverage that is arguably more strategically valuable than a dozen branch fabs producing chips on mature nodes.

This approach would mean accepting that Europe will not manufacture twenty per cent of the world's chips by 2030 — or probably ever. It would mean accepting that European automakers will continue to source some of their chips from Asian fabs, just as Asian smartphone manufacturers source their lithography tools from the Netherlands. It would mean redefining sovereignty not as self-sufficiency in production but as indispensability in the supply chain — the assurance that no one can build advanced chips without European technology, just as no one can currently build them without American design software or Taiwanese manufacturing expertise.

Sovereignty in semiconductors does not mean making every chip yourself. It means ensuring that no one can make the chips that matter without you. Europe already has this power in equipment and research. The question is whether it recognises what it has — or squanders it chasing a production target it cannot reach.

Editorial analysis

The Dresden fab is not a mistake. European automakers genuinely need local supply of mature-node chips, and a TSMC-operated facility in Saxony is a credible way to provide it. The European Chips Act is not a failure. Its research investments, though modest, are directed at areas of genuine European strength. But neither the fab nor the Act constitute semiconductor sovereignty in any meaningful sense. They are accommodations — pragmatic, defensible accommodations — to a global industrial architecture in which Europe plays important but specific roles, none of which include the large-scale manufacture of the most advanced chips.

The danger is not in making these accommodations. It is in mischaracterising them. When European leaders announce a TSMC branch fab as a sovereignty milestone, they create expectations that the project cannot fulfil and obscure the strategic investments that might actually matter. When the European Commission sets a twenty per cent production target that industry analysts consider unreachable, it substitutes aspiration for analysis and invites the inevitable narrative of failure when the target is missed. When billions in subsidies flow to foreign-owned fabs while indigenous research programmes receive a fraction of the funding, the revealed priorities contradict the stated strategy.

Europe's semiconductor future will be built on what Europe is actually good at: the tools that make chips possible, the research that defines what chips will become, and the specialised chip design that serves the industries — automotive, industrial, energy — where European companies lead globally. Dresden will produce chips. It will not produce sovereignty. The sooner European policymakers internalise this distinction, the sooner they can build a semiconductor strategy worthy of the continent's genuine capabilities.

Sources

1. TSMC European fab announcement — https://pr.tsmc.com/english/news/3049
2. European Chips Act — Official Journal of the EU — https://eur-lex.europa.eu/eli/reg/2023/1781/oj
3. Intel Magdeburg fab pause — Reuters — https://www.reuters.com/technology/intel-pauses-construction-germany-chip-plant-2024-09-16/
4. ASML Annual Report 2023 — https://www.asml.com/en/investors/annual-report
5. Semiconductor Industry Association — Global Fab Capacity — https://www.semiconductors.org/turning-the-tide-for-semiconductor-manufacturing-in-the-u-s/
6. imec — Semiconductor R&D Centre — https://www.imec-int.com/en/about-us
7. Volkswagen chip shortage production impact — https://www.reuters.com/business/autos-transportation/volkswagen-produced-600000-fewer-vehicles-2021-due-chip-shortage-2022-01-12/
8. European Court of Auditors — EU Chips Act Assessment — https://www.eca.europa.eu/en/publications/SR-2024-20