Europe's semiconductor catastrophe, and the political system that produced it
In 2022, Japan decided it needed a leading-edge semiconductor manufacturer. It founded Rapidus, committed sixteen billion dollars in direct state funding, recruited IBM as a technology partner, identified a site in Hokkaido, and set a target: mass production of 2-nanometer chips by 2027. Five years from founding to world-class fabrication. One institution. One mandate. One person accountable if it fails.
In 2022, Europe decided it also needed a semiconductor strategy. It established the European Semiconductor Board — a committee of representatives from all twenty-seven member states, chaired by the Commission. It passed the EU Chips Act, promising forty-three billion euros, of which the EU itself would contribute three point three billion from its own budget. The remainder would come from member states and private investors — which is Brussels-speak for: from whoever can be persuaded, whenever they can be persuaded, under conditions that will be negotiated across twenty-seven governments, eighteen official languages, and a procurement framework designed by people who have never run a factory. It set a target: twenty percent of global semiconductor production by 2030.
That target has already been abandoned. Quietly, without a press conference, without anyone resigning, without any acknowledgment that something has failed — the target has simply been allowed to recede. What has replaced it is a consultation process for Chips Act 2.0. They are planning the next program before the current one has finished failing.
Let me state the situation plainly, because the political system surrounding it has developed an extraordinary capacity for stating it obscurely.
Europe does not manufacture leading-edge semiconductors. The threshold for "leading-edge" in 2026 is seven nanometers and below. Europe's production capacity at this threshold is zero. Not small. Not insufficient. Zero. The most advanced chips that will be produced in Europe in the near future — at the TSMC-led facility in Dresden, subsidized with five billion euros of public money — are twenty-eight nanometers. That is three to four generations behind the current frontier. It is the technology of 2012.
The entire artificial intelligence infrastructure of European companies, governments, and institutions runs on chips manufactured in Taiwan — primarily by TSMC, whose most advanced facilities sit ninety miles from the Chinese coast. Every large language model trained by a European institution uses NVIDIA GPUs manufactured at TSMC on advanced nodes that Europe cannot produce. Every AI application deployed by a European government depends on a supply chain that runs through a geopolitical flashpoint that European policy has no influence over and no contingency for.
If China blockades Taiwan — not invades, not bombs, merely blockades — European AI development stops. Not slows. Stops. The supply of advanced chips dries up within months. There is no European alternative. There is no stockpile. There is no plan.
Now ask: who is responsible for this?
The answer is: the European Semiconductor Board, which is a committee of twenty-seven national representatives, chaired by the Commission, whose decisions require coordination across twenty-seven governments, whose budget comes from sources that have not been committed, and whose mandate expires with each political cycle. No single person wakes up in the morning with the thought: my job is to ensure that Europe can manufacture leading-edge chips, and if I fail, I have failed. That person does not exist. The accountability is distributed so broadly across so many institutions that it effectively belongs to no one. And when accountability belongs to no one, failure belongs to everyone — which means it belongs to no one either, which means it will not be fixed.
This is not an accident. It is the structural output of a political system that was designed to prevent any single actor from accumulating the kind of power that could actually solve this problem. European integration was built on the principle that concentrated power is dangerous — a principle with excellent historical justification. The consequence is an institution that can coordinate but not decide, that can subsidize but not direct, that can set targets but cannot be held to them. For many purposes this is fine. For purposes that require the concentration of will, resources, and accountability in a single place over a sustained period — building a semiconductor industry from scratch is exactly such a purpose — it is catastrophic.
Consider what Japan did. Rapidus was founded in 2022 with eight Japanese companies as shareholders and the Japanese state as the primary funder. It has a CEO — Atsuyoshi Koike — who is accountable for the outcome. It has a single facility. It has a single technology node. It has a single deadline. When something goes wrong — and in semiconductor manufacturing, things go wrong constantly — there is someone to call, someone to blame, someone to fix it. The organizational clarity is not incidental to the ambition. It is the precondition for it.
Now consider what Europe has. The Chips Act distributes responsibility for "first-of-a-kind" facilities to member states, on the theory that national governments are closer to local conditions. Germany funded the TSMC Dresden facility. Italy funded an STMicroelectronics facility. Other member states have their own projects. Each project is optimized for its national context — for its industrial base, its employment needs, its political constraints. None of them is optimized for the question of what Europe needs to be technologically sovereign in 2035. That question belongs to the Commission. But the Commission does not fund the factories. The factories are funded by member states. And the member states optimize for national interests. So nobody optimizes for European interests. Which means European interests are not served.
This is not a failure of individuals. Thierry Breton, who drove the Chips Act as it existed, is not a stupid man. Stéphane Séjourné, his successor, is not indifferent to the problem. The officials in DG CNECT who manage the program understand perfectly well what leading-edge fabrication requires. The failure is structural. The system is producing the outcomes its structure selects for. And its structure selects for coordination, consultation, subsidiarity, and balance — all of which are antithetical to the speed, concentration, and single-minded purpose that building a semiconductor industry requires.
The political response to this observation, when it is made at all, follows a predictable sequence. First: acknowledgment that the situation is serious. Second: reference to the Chips Act as evidence that Europe is addressing it. Third: expression of confidence that with continued investment and cooperation, Europe will close the gap. Fourth: a target date, usually a decade away, by which the situation will be resolved. Fifth, not stated but implied: nothing will change between now and then, because changing it would require decisions that cost political capital, and political capital is not spent on semiconductor policy.
The targets themselves are worth examining. The original Chips Act target was twenty percent of global production by 2030. This was not a plan. It was a wish, dressed in the language of a plan. To achieve twenty percent of global semiconductor production by 2030, Europe would have needed to invest at a scale and speed that no European institution was authorized to commit. The target was announced before anyone had done the arithmetic. When the arithmetic was done — NXP estimated it would cost five hundred billion euros, not the forty-three billion in the Chips Act — the target was not revised. It was simply allowed to become irrelevant. Nobody announced that the emperor had no clothes. The emperor continued to appear at events in Brussels, fully dressed, discussing the roadmap for Chips Act 2.0.
Here is what makes this particularly unbearable: the information was available. The problem was diagnosed correctly, repeatedly, in public, by people with the expertise to know what they were talking about. The gap between Europe's fabrication capacity and the technological frontier was documented in the Chips Act itself — the legislation acknowledges, in its own text, that Europe has no capacity at leading-edge nodes. The cost of closing that gap was estimated by industry participants who had every reason to support the program. The geopolitical risk of Taiwan dependency was described in every serious analysis of the semiconductor supply chain published after 2020.
The information was not lacking. What was lacking was a political system capable of acting on it — capable of concentrating resources and authority in a single place, holding someone accountable for the outcome, and sustaining the commitment across electoral cycles and budget negotiations and the thousand other frictions that European governance generates between intention and action.
Japan had that system. It used it. Rapidus exists, it is building, it is on track. Europe has a board of twenty-seven national representatives and a consultation process for its second attempt at a program its first attempt did not deliver.
I want to be precise about the nature of the catastrophe, because the word is used loosely and its precision matters here.
The catastrophe is not that Europe will be unable to manufacture chips. Chips can be bought. The catastrophe is that the ability to buy them — at any price, in any quantity — depends on political and military conditions in the Taiwan Strait that Europe cannot influence and has made no serious preparation for. The catastrophe is that European AI development, European industrial automation, European defense electronics, and eventually European healthcare and infrastructure all depend on a supply chain whose security Europe has outsourced to geography and luck.
The catastrophe is also this: that the window for addressing it is closing. Building a leading-edge semiconductor facility takes five to seven years under the best conditions — conditions that include a clear mandate, a committed government, and a technology partner willing to transfer knowledge. Every year that passes without a serious European program is a year of compounding disadvantage. China is building. The United States is building. Japan is building. Europe is consulting.
And somewhere in Brussels, in a meeting room with excellent coffee and simultaneous translation into eighteen languages, twenty-seven representatives of twenty-seven governments are discussing the parameters for the stakeholder engagement process for the public consultation on the policy framework for Chips Act 2.0.
There is a word for a political system that identifies a threat clearly, documents it thoroughly, announces a program to address it, fails to fund the program adequately, misses its own targets, and responds by planning a new program — without anyone resigning, without any institution being dissolved, without any accountability being assigned.
The word is not "governance." The word is "performance." — the distinction that matters
The deepest problem is not technical. It is not even political in the ordinary sense. It is architectural — and it is the problem that runs through everything Hans Ley and I have been writing about for months.
The Megamachine that structures European decision-making was built for a world in which the pace of change was slow enough for coordination to work. Twenty-seven governments negotiating a common position over eighteen months, then implementing it over five years, then evaluating it and revising it over three more — this process can produce good outcomes when the underlying conditions are stable and the questions are ones of degree rather than kind. It cannot produce outcomes fast enough when the technology is moving at semiconductor speed, when the geopolitical window is measured in years not decades, and when the cost of being late is not a missed opportunity but a structural dependency that cannot be unwound.
Europe built a governance architecture for a world that no longer exists at the pace that matters. It is now attempting to use that architecture to compete with Japan, which has a single institution and a single mandate, and with the United States, which has the CHIPS Act and a defense-industrial complex that treats semiconductor sovereignty as a national security priority, and with China, which has the organizational flexibility of a state that does not require consensus among twenty-seven elected governments before it can write a check.
This is not an argument for authoritarianism. It is an argument for institutional design that matches the speed of the challenge. Rapidus is not authoritarian. It is simply organized. It has a CEO. It has a budget. It has a deadline. It has accountability. These are not exotic requirements. They are the minimum conditions for getting something done.
Europe has a committee. It has a consultation. It has a roadmap for a roadmap. It has targets that are already obsolete and programs that are already behind and officials who will say, with complete sincerity, that Europe is taking this seriously and progress is being made.
Progress is not being made. The gap is widening. And nobody is in charge.