On 14 July 2026 the European Commission approved four more grants for chip plants in Germany and, in the same release, put the chip aid it had approved so far at around €14.2 billion.
Europe is not building its next chip plants for the race to the smallest transistor: every plant the Commission has funded since the start of 2025 makes power chips, specialty chips, materials or equipment, the one leading-edge fab exists only as a proposal, and the jobs this build creates are in power, specialty process and the optical links of data centres.
This is the third piece of the CareerOn Industry Atlas on semiconductors. The first asked where the money is made in a chip. The second asked which suppliers a chip plant could not replace. This one asks the question a candidate asks first: what is actually being built in Europe, where, and when does each plant say it will run? We answer it only from dated documents published by the European Commission and by the companies themselves. Where a document gives no date, we say so. We do not supply one.
That rule matters more in this piece than in any other. Construction news travels fast and loosely. A topping-out ceremony becomes a production start, a projected cost becomes an investment, a proposal becomes a policy. The documents are more careful than the headlines, and the gap between the two is where readers get misled about where the work will be.
The Commission’s own count
When the Commission approves State aid for a chip plant, its press release ends with a running total of the aid it has approved for the sector. The number is printed by the institution that approved it, which makes it the cleanest measure available of how much public money Europe has committed to chip plants on paper.
On 21 November 2025 the Commission put that total at around €12.8 billion. On 11 December 2025 it printed around €13.2 billion. On 14 July 2026 it printed around €14.2 billion.
Three readings are not a trend line, and we do not draw one. Each figure carries the word “around” in the original, and we keep it. We do not add later decisions to the last total, we do not guess what lies between the dates, and we do not project where it goes next. What the series does show is that the approvals kept coming through a period when much of the public discussion about European chips was about projects that stalled.
The Commission’s own count of approved chip aid rose by €1.4 billion in eight months
- 21 Nov 2025: around €12.8bn cumulative approved State aid for chip facilities, as printed by the Commission, EUR billion
- 11 Dec 2025: around €13.2bn cumulative approved State aid for chip facilities, as printed by the Commission, EUR billion
- 14 Jul 2026: around €14.2bn cumulative approved State aid for chip facilities, as printed by the Commission, EUR billion
cumulative approved State aid for chip facilities, as printed by the Commission, EUR billion
Sources 2 European Commission · 3 European Commission · 7 European Commission
What the money builds
The more useful question is not how much, but what for. We read the decisions the Commission issued between February 2025 and July 2026 plant by plant, and recorded what each one says the plant will make.
On 20 February 2025 the Commission approved a direct grant of up to €920 million for Infineon’s new plant in Dresden, an investment of €3.5 billion. The decision describes two kinds of product: technologies used for power switching, management and control in electronic systems, and analog/mixed-signal integrated circuits. It is also one of only two decisions in our set that prints a date for the plant itself, which will reach its full capacity in 2031.
On 21 November 2025 the Commission approved about €450 million for Onsemi, toward an investment of €1.64 billion in Rožnov pod Radhoštěm, in Czechia. The plant is described as a novel integrated manufacturing plant for silicon carbide power devices, and the decision says it is expected to start commercial operations by 2027.
On 11 December 2025 the Commission approved €495 million for GlobalFoundries in Dresden and €128 million for X-FAB in Erfurt. GlobalFoundries is described as a pure-play foundry, making chips exclusively for other companies, and its project is to be adapted for dual-use purposes, aimed at aerospace, defence and critical infrastructure. X-FAB’s plant is a new open foundry that will provide manufacturing services to fabless chip companies, including start-ups and smaller firms.
On 14 July 2026 came four more German decisions: €353 million for Element 3-5 in Baesweiler, for silicon carbide epi-wafers; €214 million for Vishay in Itzehoe, for power MOSFETs; €74.4 million for KLA in Weilburg, for metrology equipment; and €17.9 million for KETEK in Munich, for detector chips.
Put the eight plants side by side and a pattern appears that no single announcement makes plain. Four make power devices or the materials they are built on. Two are foundries serving other companies’ designs, one of them reoriented toward defence and critical infrastructure. One makes the instruments that measure chips during manufacture. One makes specialised detectors. None of them makes the most advanced processors, the chips that dominate the headlines about artificial intelligence.
Eight plants funded since 2025, and none of them makes leading-edge logic
| Plant | What it makes | Approved aid | Runs from, as printed |
|---|---|---|---|
| Infineon, Dresden | power and analog/mixed-signal chips | up to €920m | full capacity in 2031 |
| Onsemi, Rožnov pod Radhoštěm | silicon carbide power devices | about €450m | commercial operations by 2027 |
| GlobalFoundries, Dresden | foundry capacity adapted for dual use | €495m | not printed |
| X-FAB, Erfurt | an open foundry for fabless companies | €128m | not printed |
| Element 3-5, Baesweiler | silicon carbide epi-wafers | €353m | not printed |
| Vishay, Itzehoe | power MOSFETs | €214m | not printed |
| KLA, Weilburg | metrology equipment | €74.4m | not printed |
| KETEK, Munich | detector chips | €17.9m | not printed |
Sources 1 European Commission · 2 European Commission · 3 European Commission · 7 European Commission
Only two of the eight decisions give a date for the plant: Infineon at full capacity in 2031, Onsemi in commercial operation by 2027. For the other six, the date column in our table says “not printed”. That is not a gap in our research. It is what the documents say, and a reader planning a move deserves to know that a date circulating in the press may have no source behind it.
The leading edge is a proposal
If Europe is not building leading-edge plants now, is it planning to? The clearest answer comes from the Commission itself. On 3 June 2026 it published its proposal for a revised Chips Act, known as COM(2026) 504. The proposal states that the Union produces less than 10% of the world’s semiconductors and is almost entirely dependent on the United States and Asia for the most advanced chips, those below 5 nanometres.
It then sets out the ambition: to establish the first facility in the Union combining leading-edge node manufacturing with chiplet integration and advanced 3D packaging. And it gives a horizon in a single sentence: pilot production could be envisaged in the period 2030-2033.
Three words in that sentence carry weight. It is a proposal, not law: the Parliament and the Council have yet to adopt it. It speaks of pilot production, not volume. And “could be envisaged” is the language of possibility, not of commitment. We print it as the Commission wrote it, as a proposed horizon for a plant that does not yet exist.
Dresden’s other fab
The obvious objection is TSMC. The world’s largest contract chipmaker is building a plant in Dresden with Bosch, Infineon and NXP, through a joint company called ESMC, and it is often described as Europe’s advanced fab. TSMC’s own words are more precise. In its annual report for 2025, published in April 2026, it writes that it is building “a specialty technology fab in Dresden, Germany, focusing on automotive and industrial applications.”
Specialty, automotive, industrial: the company doing the building describes the plant in the same terms as the rest of the list. It is a large and serious project, and it will employ many people. It is not a leading-edge logic fab, and its builder does not say it is.
We do not print a production year for this plant. Press reports disagree on when production starts, and the document that would settle it, TSMC’s own quarterly statement, could not be opened by our source checks. Until a dated document from TSMC or ESMC prints a year, we leave the year out.
Crolles: the plan and the record
The plant closest to our readers in Isère shows why dates deserve suspicion. On 11 July 2022 STMicroelectronics and GlobalFoundries announced a new joint facility next to ST’s existing plant at Crolles, near Grenoble. The release was specific: the facility was “targeted to ramp at full capacity by 2026, with up to 620,000 300mm wafer per year production at full build-out”.
ST’s annual filing for 2025, dated 26 February 2026, still carries the project. It gives the projected cost as €7.5 billion and the French support as up to roughly €2.9 billion, and it records that ST received around €126 million in cash and recognised grants of €72 million from the programme in 2025. What it does not print, anywhere we could find in its Crolles passages, is a ramp date. The target of full capacity by 2026 is not repeated.
Crolles in 2022 promised full capacity by 2026, and the 2026 filing prints no date
| Document | What ST wrote |
|---|---|
| Press release, 11 Jul 2022 | This facility is targeted to ramp at full capacity by 2026, with up to 620,000 300mm wafer per year production at full build-out |
| Form 20-F, 26 Feb 2026 | 300mm wafer fab evolution for digital and Cloud Optical Interconnect |
Sources 8 STMicroelectronics · 6 STMicroelectronics
We draw no conclusion from that silence, and neither should a reader. A filing that stops restating a target is not the same as a company announcing a delay, and we will not write “missed” or “delayed” unless ST or GlobalFoundries does. What the filing does say is where ST wants to take the site. Its capital plan describes “300mm wafer fab evolution for digital and Cloud Optical Interconnect.”
Where artificial intelligence reaches a European fab
That phrase is easy to pass over, and it may be the most important one in this piece. Cloud optical interconnect means the optical links that carry data between machines inside data centres. As the compute buildings we described earlier in the year grow, the bottleneck moves from the processors to the connections between them, and those connections increasingly use light instead of copper.
Our reading, which the documents support but do not state, is this: Europe does not need to make the most advanced processors to take part in the build-out of artificial intelligence. It can take part through the parts around them. Power chips feed the buildings. Specialty foundries make the controllers and sensors. Optical interconnect carries the data. The Crolles phrase is the one place in our documents where a European chipmaker names that link directly.
The counter-case
The strongest objection to our reading is that it describes the present and mistakes it for a strategy. Europe is funding specialty plants now because they are the plants companies asked to build, and the revised Chips Act exists precisely because the Commission wants something more. If the proposal is adopted and a leading-edge pilot line is built, the picture in this piece will be out of date.
That objection is fair, and it is why we print the Commission’s horizon as a horizon. But a candidate choosing a first job today works in the plants that exist or are being built today, not in one proposed for the next decade. The second objection is that a list of approved grants is not a list of working plants: approval is not construction, and construction is not production. That is also true, and it is why our table shows only the dates the decisions print.
What would change the reading
Three things would change it. A decision by the Parliament and the Council adopting the revised Chips Act, with funding and a site for the leading-edge facility. A dated document from TSMC or ESMC that sets a production year for Dresden and describes the process it will run. And a filing from ST or GlobalFoundries that restates, revises or withdraws the Crolles target. We will read each one when it is published, and we will update this piece in the open.
What this means for your career
If the thesis holds, the European chip jobs of the next few years are concentrated in a few families, and none of them requires a leading-edge logic fab. Three of them can be rehearsed on CareerOn today.
The first is process integration on specialty technologies, the kind of work Crolles has built its reputation on. Our FD-SOI process integration simulation puts you in charge of bringing a process flow together so that the chip works at the end of it.
The second follows the phrase in ST’s filing. Our silicon photonics transceiver simulation asks you to design the optical link that carries data between machines in a data centre, the point where artificial intelligence meets a European fab.
The third follows the power and automotive plants that fill most of the table. Their chips go into cars, grids and factories, and none of them ships until it has passed qualification. Our automotive test and qualification simulation asks you to take a chip through that process.
None of these simulations will tell you which plant to join. What they can show you, in an afternoon, is which of these jobs you would want to do for years, and that is a better guide than any headline about the smallest transistor.