What stays when the subsidy ends

The premise

If what a plant leaves behind is counted only as buildings and headcount, that count will not show where the design, the equipment and the judgement end up once the subsidy stops.

That’s a condition, not a conclusion. This piece doesn’t say subsidies are wrong, or that the plants aren’t needed.

Two plants in Dresden

Two chip plants are going up in Dresden.

One is ESMC, built by Taiwan’s TSMC together with Bosch, Infineon and NXP. The investment runs to more than €10 billion, and at full capacity the plant will turn out 40,000 wafers a month on 28/22 and 16/12 nanometre processes. The partners put direct employment at about 2,000 jobs.

The other is Infineon’s Smart Power Fab: €5 billion for power semiconductors, and up to 1,000 new jobs, by the company’s own account.

These are large plants. On the headcount, though, they work out at roughly one job per €50 million invested. That division is ours and appears in no official document, but it gives a sense of scale. A car plant with the same money behind it would have employed far more people.

Which raises the question. When the subsidy ends, what kind of work is left in Europe?

Europe is asking a version of it too. On 24 September, EU competitiveness ministers held a policy debate — not conclusions — on the next chips act. The presidency’s note on the meeting put it like this:

“The next chips act must turn ambition into real sovereignty. In the global race for AI and advanced technologies, Europe cannot afford to be just a customer.”

Not a customer, then. So what does being a producer consist of?

Target and forecast

The European Court of Auditors reported on the record so far this year.

The target is 20% of world production by value by 2030, in cutting-edge and sustainable semiconductors.
The money attached to it is put at at least €86 billion. Opened up, that is €43 billion of policy-driven investment plus a matching €43 billion that private investors are expected to add. Half of the total is an expectation.
And the Commission’s own forecast, published in July 2024, is 11.7% by 2030 — a little over half the target.
The same report cites a 2022 ASML position paper: reaching a 20% share would call for capital expenditure of $264 billion, about €251 billion, by 2030.

The auditors’ verdict is in their subtitle: reasonable progress in implementation, but the Chips Act is very unlikely to be enough for the target.

Set the figures side by side and it reads: €251 billion cited as the requirement, €86 billion identified, and €43 billion of that actually public money.

Where Europe already stands

The same report maps the supply chain stage by stage, naming who sits where.

Stage	Names in the report
Design	US (Qualcomm, AMD), Taiwan (TSMC)
Equipment	EU (ASML), US (Applied Materials), Japan (Tokyo Electron)
R&D	US (Intel, NVIDIA, Qualcomm), EU (ASML, IMEC)
Wafer fabrication	Taiwan (TSMC), South Korea (Samsung), China
Assembly, test, packaging	Southeast Asia (Malaysia, Viet Nam, China)
Raw materials	Japan (50% of global silicon wafer supply)

Europe’s name appears under equipment and R&D. It does not appear under wafer fabrication. The Commission’s own summary of the new proposal points the same way: the EU, it says, remains dependent on third countries in key areas such as advanced chip manufacturing and semiconductor design.

The Dresden plants fill in one of those boxes. The boxes Europe is already strong in are different ones. Which means that “semiconductor jobs” needs unpacking before it means anything: jobs baking wafers, jobs building the machines that bake them, or jobs designing what goes on them.

Wages, or running costs?

That European plants are expensive is a familiar line. What exactly is expensive differs by country.

For the second half of 2025, industrial electricity (annual consumption of 2,000–20,000 MWh, excluding VAT) cost €0.1922 per kWh in Germany, €0.1596 across the EU and €0.1571 in Belgium. Gas came to €0.0713, €0.0605 and €0.0481. Germany is above the average; Belgium sits around it.

Labour taxation runs the other way. For a single worker on the average wage, taxes and social contributions take 52.5% of labour costs in Belgium, 49.3% in Germany and 30.0% in the United States. That money doesn’t vanish: it pays in advance for healthcare, pensions and unemployment cover, which in the US is partly paid separately by individuals and private insurers.

So “Europe is expensive” doesn’t survive as one sentence. In Germany the weight is on electricity; in Belgium it is on labour. And which of them decides anything depends on the plant: labour-heavy assembly answers to wages, capital-heavy fabs answer to power and utilisation.

There is no place where this is counted

The question here was what kind of work stays. Looking for evidence, we ran into a gap.

We found no official statistics on how the people inside a chip plant divide by occupation — how many operators, how many maintenance technicians, how many engineers. Nothing counts fabs on their own, in Europe or elsewhere. The size of the money used to attract a plant is written down in document after document; the place where the kind of work it brings would be counted is empty.

What exists is the company’s phrase, “2,000 high-tech professional jobs.” The phrase doesn’t say what those 2,000 people do.

After the subsidy

Intel agreed with the German government in 2023 to build two plants in Magdeburg, an investment then put at more than €30 billion. In its second-quarter results for 2025, the company wrote that it will no longer move forward with the planned projects in Germany and Poland, as part of its cost reductions.

Attracting a plant and keeping one are different achievements. Which gives a testable question: once the subsidy ends, can the plant fund its next generation of equipment from Europe?

That cannot be answered yet. ESMC is aiming to start production by the end of 2027; Infineon’s new fab starts up this autumn. Who pays for the equipment after that becomes visible later.

What can be separated now is the purpose. Plants are attracted for more than one reason: to create jobs, to hold the ability to make things when supply is cut off, to keep core technology and decisions inside, to protect tax revenue and a local economy, to raise productivity. Those five are different goals, and they are not measured by the same number.

By headcount, a chip plant is small. But headcount is one of the five. Unless you say which one you are measuring, the same plant can be called a success and a failure with equal confidence.

What this piece does not claim

That the subsidies are a waste
That Europe has lost the race
That regulation should be loosened or taxes cut
That immigration or automation is the answer
That these plants will or will not outlast the subsidy

What we checked, and what we couldn’t

Checked

ESMC Dresden: more than €10 billion, about 2,000 direct jobs, 40,000 wafers a month, 28/22 and 16/12 nanometre — TSMC announcement
Infineon Dresden: €5 billion, up to 1,000 new jobs — company materials
The Competitiveness Council of 24 September 2026 held a policy debate rather than adopting conclusions, and the presidency’s note carries the line about Europe not being able to remain just a customer
The 20% target is 20% of world production by value in cutting-edge and sustainable semiconductors by 2030 — ECA special report 12/2025
The €86 billion is €43 billion in policy-driven investment plus an expected matching €43 billion in private investment — same report
The 11.7% projection for 2030 is the Commission’s own forecast of July 2024 — same report
ASML’s 2022 estimate that a 20% share would require $264 billion (about €251 billion) — cited in the same report
In the report’s supply chain map, Europe is named under equipment and R&D — same report
The Commission’s summary of its 3 June 2026 proposal says the EU remains dependent on third countries in advanced manufacturing and chip design
Intel’s second-quarter 2025 results say it will no longer move forward with its German and Polish projects; the 2023 agreement covered more than €30 billion at Magdeburg
Industrial electricity and gas prices for the second half of 2025 — our own calculation from Eurostat raw data
Tax and social contributions as a share of labour costs in 2025: Belgium 52.5%, Germany 49.3%, United States 30.0%

Couldn’t check

The text of the new proposal
The priority areas ministers named
The Crolles state aid decision and its employment plans
Occupational statistics inside chip plants
The size of Europe’s shortage of semiconductor workers
The division of investment by headcount is ours
Whether these plants will fund their next equipment once the subsidy ends