For twenty years automotive was the core of European engineering: Volkswagen, Bosch, Continental, Renault, Fiat were, throughout this period, the reference names of the sector, from the assembly line to the R&D department. That primacy had never been called into question, not even during previous crises.
The most recent data point to a reversal of this trend.
Analysis of more than 3,400 corporate restructuring events recorded by Eurofound's ERM database between 2002 and 2026, across 29 European countries (EU-27, Norway and the UK) and seven engineering segments, produces a clear picture: automotive is going through a structural crisis with no precedent in the observed period, while other sectors — defense foremost — are progressively taking its place as the employment engine of continental engineering. As shown further below, it is far from a given, however, that artificial intelligence — the sector receiving most of today's attention — can play the same role.
The two lines in this chart tell the story better than any headline: for twenty years automotive oscillated above and below zero following the economic cycle, while defense remained background noise, almost always slightly negative. From 2022 onward, the two have literally swapped places.
The scale of the automotive contraction
The numbers speak for themselves. Over the full 2002–2026 period, the net automotive employment balance across the 29 countries covered is −272,118 jobs — the worst figure among the seven engineering sectors tracked. A significant point concerns its distribution over time: almost half of that figure (−135,425 jobs, 49.8% of the total) is concentrated in the last five years, between 2022 and 2026.
This is not, therefore, a linear decline spread evenly over time, but a recent and marked concentration.
This figure is consistent with what has been independently documented by industry sources. CLEPA, the European association of automotive suppliers, has calculated that between 2024 and 2025 component suppliers announced more than 104,000 job cuts, against roughly 7,000 new positions created over the same period. A Eurofound report on the sector's crisis, published in late 2025, shows that automotive employment as a whole (production, sales, aftermarket) grew 12% between 2011 and 2023, but this aggregate figure masked losses of more than 7% in manufacturing and supplier segments alone over the same span, with the decline concentrated above all in France, Germany and Italy. Even Central and Eastern European countries, which until 2023 had seen automotive employment grow by absorbing production relocated from Western countries, have themselves begun losing jobs, as production has shifted further toward lower-cost non-EU countries.
The German case and the role of the Volkswagen group
Breaking the figure down by country, Germany turns out not simply to be the hardest-hit country, but to represent 77% of all European automotive losses recorded between 2022 and 2026 (−104,320 jobs out of a European total of −135,425). No other single country comes close to this figure.
A more precise reading, however, requires breaking the figure down further by company, because the term "German decline" imprecisely describes what the data show: a single industrial galaxy accounts for most of the drop. Volkswagen, together with Audi and Porsche, alone accounts for more than 40% of all German automotive job losses over the last five years — a value roughly equivalent to that of all the country's other companies combined, including long-standing suppliers such as Bosch, ZF Friedrichshafen, Continental and Schaeffler, which work largely for the Wolfsburg group itself.
The most significant component of this figure is a single event: the "Zukunft Volkswagen" agreement, signed on 20 December 2024 between the group, the IG Metall union and the works council. It is a plan for 35,000 fewer jobs by 2030, alongside parallel agreements for roughly another 7,000 jobs at Audi and nearly 2,000 at Porsche. One point needs clarifying, because it affects how the figure should be read: these are not immediate layoffs. The agreement explicitly rules out forced redundancies and plant closures through 2030, and provides for the reduction to happen through early retirement, non-replacement of employees who leave voluntarily, and exit incentives. In exchange, workers accepted a wage freeze and a production capacity cut of about 734,000 vehicles a year — roughly the entire output of the Wolfsburg plant.
2022 and 2023 show an overall positive engineering balance (+11,200 and +11,389 jobs respectively); 2024 instead drops to −70,786, a decline explained almost entirely by the single Volkswagen announcement. Before that event, European engineering as a whole showed a relatively positive trend. This is therefore not a continuous aggregate decline, but a single large-scale event that explains the drop in one specific year — a distinction worth keeping in mind whenever an isolated spike appears in this data.
The sectors that are growing, and the ones lagging despite the announcements
The automotive contraction is matched by growth in other segments. Defense/Aerospace shows the sharpest reversal: from a cumulative negative balance of −45,391 jobs over the full 2002–2026 period (including the last five years), to a positive balance of +57,837 jobs in the 2022–2026 period alone — the only segment, among the seven tracked, to fully flip sign. The figure is consistent with what has been documented at sector level: in 2024 European defense revenue grew 13.8% to €183.4 billion, while direct employment rose 6.9% to 1,103,000 jobs, the highest level ever recorded, driven by the rearmament launched after Russia's invasion of Ukraine in 2022.
Three other sectors show the same turnaround, on a smaller scale: Construction/AEC (+12,872 jobs in 2022–2026), Energy (+10,581) and Rail/Naval (+5,913). None of these come close to defense's volumes, but the direction is the same. Dedicated deep dives: Defense/Aerospace sector page and Energy sector page.
Electronics/Semiconductors: a sector that has not seen the expected growth
A notable case worth reporting for completeness is Electronics/Semiconductors, which remains negative even in 2022–2026 (−11,213 jobs), despite the public investment announced under the European Chips Act. This number is also likely to understate the real scale of the problem.
The reason is structural to how ERM works: the database records restructurings of existing employment, not the cancellation of future hiring plans that were never launched. A notable case is that of the €30 billion mega-fab Intel had announced in Magdeburg, Germany, together with a satellite facility near Wrocław, Poland — a project that would have brought about 21,000 direct and indirect jobs, intended for producing latest-generation chips (sub-5nm nodes, the same used for artificial intelligence). The project was officially cancelled in July 2025, for lack of sufficient customer orders. Those promised 21,000 jobs that were never created do not appear in this dataset, since they never existed as employment to be restructured. The gap between European expectations for the semiconductor supply chain and the results actually achieved is therefore wider than these numbers, on their own, can show.
The only major project still moving forward — the TSMC-Bosch-Infineon-NXP joint venture in Dresden — targets mature production nodes (28/22nm, 16/12nm) for automotive and industry, not AI chips. The European Court of Auditors has already declared it "highly unlikely" that the EU will reach its target of a 20% global market share by 2030 (realistic projection: about 11.7%).
The lag in the AI supply chain is not the main explanation
One possible reading of European engineering's problem is that the delay in AI development is due to slow and fragmented political execution — a hypothesis partly confirmed by the Intel episode. This is, however, an incomplete reading: even if Europe had built its advanced chip production capacity in time, the employment numbers at the companies leading AI development — Nvidia, OpenAI, Anthropic — would still be unable to offset automotive's employment decline. There are at least three reasons for this, structural in nature rather than reflecting a reversible policy choice.
The first concerns skill adjacency, and is why the automotive-defense transition is more workable than a hypothetical automotive-AI transition. A mechanical engineer who has worked on engines, transmissions and active safety systems in a car has a technical base — systems mechanics, materials, embedded electronics — that transfers reasonably well to a military aircraft or a naval platform: it is the same type of engineering, physical and safety-critical. The move toward the AI supply chain, by contrast, is different: chip design is highly specialized electronic engineering, while model training is largely software engineering and applied mathematics, disciplines that share little with powertrain mechanics. The one area of real overlap (power electronics for batteries, transferable to data-center power supply) concerns a specific niche, not the bulk of the departing automotive workforce.
The second reason is economic: AI pays double or triple what classic engineering sectors do — an AI/ML engineer in Germany earns on average €72–92 thousand a year, in the United States the figure rises to $147–160 thousand, and at leading labs such as Anthropic and OpenAI senior roles exceed $200 thousand — but hires in much smaller absolute numbers. The reason is not a corporate choice, but the very nature of the sector: AI is capital- and skill-intensive and scales without needing labor proportional to output — a small team builds a model used by hundreds of millions of people. Automotive manufacturing, by contrast, is labor-intensive and scales almost linearly with volumes: more cars produced require more workers, more suppliers, more quality-control staff.
The third reason is less immediate: the very AI tools in question make the work of those developing artificial intelligence more productive, reducing over time the number of engineers needed per unit of growth in the sector itself — a sector that, by its technological nature, tends to produce more with a proportionally smaller headcount.
The conclusion that emerges from these elements is that AI drives revenue and market value, but is not able to structurally replace the employment mass that automotive is shedding. Defense, by contrast, remains physical, labor-intensive manufacturing, with an employment intensity per unit of output comparable to that of the automotive sector it is partly replacing — a factor that helps explain why it is defense, not electronics, that flips sign in this data.
The comparison with the United States
A useful comparison for assessing whether this dynamic is a specifically European phenomenon or reflects broader trends comes from Eurostat data (employment by NACE activity, EU-27) and FRED/BLS (employment by NAICS sector, USA) — official sources on both sides of the Atlantic, even though conceptually different from the ERM event-based net balance, since here what is measured is the annual change in the employment stock, not restructuring announcements. A notable pattern emerges: US automotive shows much more contained swings than its European counterpart, without the sharp 2022–2024 drop observed in Germany. The most plausible explanation is not that American automotive is in better shape overall, but that it has no single vertically integrated group of Volkswagen's scale whose crisis can hit the entire national segment in one wave.
The electronics comparison also confirms, from another angle, the point discussed above about AI: US employment in semiconductor manufacturing (NAICS category 3344) has stayed essentially flat — around 370–395 thousand units from 2010 to today — despite the growth in revenue and market value of American AI companies. The reason is the same one that emerged earlier: physical production of advanced chips remains largely in Taiwan (TSMC) and Korea (Samsung), while US AI companies focus on design, a low-labor-intensity activity relative to the value it generates. Any AI-related employment growth would therefore need to be sought elsewhere: in the construction and operation of data centers, a segment that falls outside the traditional definition of "engineering" adopted in this analysis.
What this data does not say
No dataset tells the whole story, and it is more honest to state that than to leave it implied.
Geographic coverage. The ERM database covers the 27 EU member states, Norway and the UK — not Serbia, where Stellantis has a strategic plant in Kragujevac, nor other non-EU European countries.
The UK stops at 2020. Following Brexit, Eurofound stopped collecting UK data after 31 January 2020. Any apparent stabilization in recent British data reflects the absence of measurements, not a genuine improvement in the employment situation.
The sector classification is heuristic. There is no official sector code attached to each event in the database's public view — the assignment to the seven engineering segments is based on company-name recognition, verified and corrected multiple times during the analysis. About 90% of the more than 31,000 total events in the database concern non-engineering sectors — retail, healthcare, financial services — and are correctly excluded from this analysis.
The questions that remain open
The numbers describe a shift in center of gravity that is already under way. It also needs to be framed within a longer horizon than the 24 years analyzed in this report: every technological innovation in history has produced temporary sectoral unemployment — from mechanical looms and the Luddites, to the mechanization of agriculture, to industrial automation in the 1980s and 1990s — and economies have historically ended up absorbing that labor elsewhere, over the long run. A plausible hypothesis is that the same will happen in this case too.
This is, however, a conclusion to be treated with caution. The cost of this transition falls disproportionately on those with the least room to reinvent themselves — a fifty-year-old line worker in Wolfsburg does not have the same range of options as a fresh graduate — and it remains an open question whether defense can genuinely absorb, in terms of skills and not just volumes, the workforce automotive is shedding. It also remains to be seen whether Germany's decline, driven so markedly by a single industrial group, is a cyclical problem tied to the delay in the electric transition, or a structural problem of the vertical-integration model on which German industry has been built over the past seventy years.
These are not questions this report can answer definitively. They remain relevant, however, for those operating directly in these sectors — for assessing a supplier chain's exposure to a single client in crisis, the opportunity of repositioning toward defense or energy, or simply for having a reliable, up-to-date reading of one's own market.
Notes
Methodology: analysis of 31,427 corporate restructuring events from Eurofound's European Restructuring Monitor (ERM) database, 2002–2026 period, 29 countries (EU-27, Norway, United Kingdom). Sector classification by company-name recognition, spot-checked. International comparison based on Eurostat data (employment by NACE Rev.2 activity) and FRED/BLS (Current Employment Statistics by NAICS), both official statistical sources. External sources cited: CLEPA, Eurofound, EY, ASD (Aerospace, Security and Defence Industries Association of Europe), European Court of Auditors.