Executive brief
- This is a forecast, and we show the arithmetic. The figure comes from a published base population (Eurostat ICT specialists), a published policy trajectory (the European Commission's Digital Decade target of 20 million ICT specialists by 2030) and an explicit growth assumption. Change the assumption and the number changes — which is the point.
- Net creation and gross hiring are different numbers. Roughly 200,000 net new roles across a year sits inside a much larger volume of gross vacancies, because replacement demand from retirements and job-to-job moves dwarfs net growth. Candidates compete in the gross market.
- The growth is not in "tech companies". The fastest-expanding technical demand is inside industrial, energy, defence and health organisations that are not classified as software firms. That is where the sovereignty build-out is spending.
- Four demand engines dominate: AI and data engineering, cybersecurity and compliance, electrification and grid software, and semiconductor and advanced-manufacturing engineering. Each is driven by a regulation or a capital programme, not by sentiment.
- Compliance became a hiring category. The AI Act, NIS2 and CSRD convert obligation into headcount. Roles that combine engineering literacy with evidence discipline are the least substitutable profiles in the market.
Almost every hiring forecast you will read this year has the same defect: a confident number with no visible method. It cannot be checked, so it cannot be wrong, so it is not information. This piece takes the opposite approach. We state the base population, the trajectory, the assumption and the arithmetic; we separate what is measured from what is projected; and we end with the observable signals that would prove the projection wrong. If you disagree with the assumption, you can substitute your own and recompute in one line.
The arithmetic, in the open
Eurostat publishes employment of ICT specialists across the EU. That series has run around ten million people in recent years, growing consistently faster than total employment. The European Commission's Digital Decade programme sets a target of 20 million ICT specialists in the EU by 2030, and its annual progress reporting has repeatedly stated that the Union is not on track to reach it.
Both facts matter. The target is what policy is spending against; the shortfall is why the spending continues. The derivation is then straightforward:
| Input | Value used | Source basis |
|---|---|---|
| Base population, EU ICT specialists | ~10 million | Eurostat employment statistics (measured) |
| Policy target for 2030 | 20 million | European Commission, Digital Decade (stated objective) |
| Observed annual growth of the series | ~2–4 % per year | Eurostat multi-year trend (measured) |
| Net new roles at 2 % on a 10 m base | ~200,000 | Arithmetic (projection) |
| Net new roles at 4 % | ~400,000 | Arithmetic (projection) |
| Growth required to hit the 2030 target | ~12 % per year | Arithmetic on the stated objective |
Read the last row carefully, because it is the honest conclusion of the exercise: the policy target is arithmetically out of reach on any recent growth rate. Doubling a ten-million-person workforce in five years would require roughly twelve percent compound annual growth, and the series has never done that. So the realistic 2026 figure is the low end — on the order of two hundred thousand net additional technical roles across the EU — and the gap to the target is not a rounding error, it is the structural reason employers keep raising wages for specific skills instead of waiting for supply.
Net creation is the wrong number to job-hunt against
Two hundred thousand net roles is a macroeconomic statistic. It is not the number of jobs you can apply to, and confusing the two leads people to the wrong conclusion about how open the market is.
Total hiring in any year equals net creation plus replacement demand: people retiring, changing employer, changing occupation or leaving the labour force. In a workforce of ten million with normal turnover, replacement demand is several times larger than net growth. Cedefop, the EU's vocational-training and skills-forecasting agency, makes this distinction explicit in its European skills forecast, and it is the single most useful correction a candidate can apply to headline hiring numbers.
The practical implication: a slowdown in net creation does not close the market, it changes who wins in it. When net growth compresses, employers stop hiring for potential and start hiring for immediately deployable skill. That is precisely when a demonstrated, evidenced capability beats a credential — the mechanism we quantified in the salary-premium analysis.
Where the demand actually sits
The most persistent error in European tech-career planning is looking for technical jobs in technology companies. The capital is elsewhere. Energy, defence, health, mobility and manufacturing organisations are hiring software, data and security profiles at scale because their own products became digital and their own obligations became regulated.
| Demand engine | What is driving it | Roles it creates | Confidence |
|---|---|---|---|
| AI and data engineering | Model deployment moving from pilot to production; the AI Act creating documentation, evaluation and monitoring obligations | Data engineers, ML platform and MLOps engineers, evaluation and safety specialists, AI governance leads | High — the obligation is legislated, not optional |
| Cybersecurity and resilience | NIS2 widening the set of in-scope entities and attaching management accountability | Security architects, incident response, OT/ICS security, GRC and audit-evidence roles | High — non-compliance carries liability |
| Electrification and grid | Grid reinforcement, storage, data-centre load growth, and the software layer that dispatches all of it | Power systems engineers, embedded software, industrial automation, energy data scientists | High — anchored in multi-year capital programmes |
| Semiconductors and advanced manufacturing | European Chips Act fabs and equipment supply chains moving from announcement to construction and ramp | Process and equipment engineers, metrology, yield analysis, industrialisation | Medium-high — timing follows fab schedules, which slip |
| Defence and space | Sustained procurement growth and multi-nation programmes with decade-long funding lines | Systems engineers, secure software, RF and signal processing, certification and safety | High for volume, gated by clearance eligibility |
| Sustainability reporting | CSRD-driven data collection, assurance and audit-grade traceability | Data engineers on non-financial reporting, LCA analysts, assurance-facing controllers | Medium — scope and timing have been actively revised |
Notice the common structure. Five of the six engines are driven by a legal obligation or a committed capital programme rather than by market optimism. That is what makes them plannable. A hiring wave produced by enthusiasm can reverse in a quarter; a hiring wave produced by a directive with a compliance date behaves like a schedule.
Geography: the same shortage, different shapes
Aggregating the EU hides the thing that determines where you should actually look. National labour markets differ in what is scarce, not just in how scarce it is.
| Market | Concentration of demand | What makes it distinctive |
|---|---|---|
| France | Defence, nuclear and energy, semiconductors, aerospace, industrial software | Sovereignty programmes plus a clearance-gated defence tier; Grenoble, Toulouse and Île-de-France are separate markets |
| Germany | Industrial automation, automotive software, power electronics, embedded systems | Deep mid-cap engineering base; software demand attached to physical products |
| Netherlands | Semiconductor equipment, photonics, high-tech systems | An equipment ecosystem with a genuinely global moat, and correspondingly narrow skill requirements |
| Nordics | Grid, storage, industrial decarbonisation, data centres | Cheap low-carbon power pulling compute and heavy industry northwards |
| Iberia | Renewables, engineering services, shared technical centres | Fast-growing delivery capability at a lower cost base, increasingly with real design ownership |
| Poland and CEE | Software engineering, security operations, R&D centres | Volume hiring plus a visible shift from cost centre to competence centre |
The corollary is that "European tech jobs" is not a market you can search. Country plus sector plus city is. The country decides the regulatory and clearance regime, the sector decides the cycle, and the city decides which specific employers are within commuting distance of the site that does the work you want.
What compresses in 2026
A forecast that only lists growth is marketing. Some categories are getting harder, and saying so is the whole basis for trusting the rest.
- Generalist junior roles. The tasks most exposed to assisted coding are precisely the tasks that used to constitute a junior brief. The WEF's Future of Jobs survey reports a large share of employers expecting significant skill disruption within five years; the first-order effect lands on entry-level task portfolios, not on senior judgement.
- Undifferentiated frontend and CRUD delivery. Still hired, but with compressed premiums and more competition per opening.
- Analytics without a domain. Dashboard production is being absorbed into tooling. Analytics attached to a regulated domain — grid, clinical, financial, safety — is not.
- Pure prompt-engineering roles. A title that existed briefly and is being reabsorbed into product and platform engineering, exactly as we argued in the AI skills paradox.
The pattern is consistent: what compresses is work whose output can be verified by inspection alone. What expands is work whose output must be defended — to a regulator, an auditor, a certification authority or a safety case. That asymmetry, not the AI capability curve, is the best available predictor of which technical roles hold their value.
How to use this if you are choosing a move
- Pick an obligation, then a company. Find the directive or capital programme that forces the hire — AI Act, NIS2, CSRD, a Chips Act fab, a defence programme — and work backwards to the employers who must comply or must build.
- Prefer regulated domains for durability. The learning curve is steeper and the resulting position is far harder to displace.
- Treat clearance as an asset class. If you are eligible, defence and critical-infrastructure work offers a structurally protected market. Start the process before you need it.
- Move up the evidence stack. Being able to prove a system behaves as claimed — tests, monitoring, documentation, audit trail — is now a differentiator rather than an overhead.
- Calibrate before you apply. Our skills diagnostic scores your profile against these demand engines and returns a prioritised reading list rather than a vanity score.
The layer the statistics miss
Eurostat counts jobs. It does not count admissibility, and admissibility is what actually decides whether a given opening is reachable by a given person. Three filters sit between the headline number and your inbox, and none of them appears in any forecast.
- Language. A large share of industrial, defence and public-sector engineering work in France and Germany is conducted in the national language, including the documentation and the audit trail. English-only candidates are competing for a materially smaller slice of the market than the aggregate suggests — and the fastest available premium for many mid-career engineers is B2 competence in the working language of the site, not another framework.
- Clearance and nationality. Defence, nuclear and critical-infrastructure roles are gated on citizenship and security vetting. This is not a preference an employer can waive, and it removes a whole tier of the fastest-growing demand from anyone ineligible.
- Site presence. Fab, plant, grid and assembly work is physically located. Remote-first candidate strategies systematically miss the categories with the strongest wage growth, because the work cannot leave the building.
The useful reframing is that these filters are not obstacles so much as pricing signals. Every filter that narrows the candidate pool raises the return to clearing it. If you can be admissible where most people cannot — the right language, the right eligibility, the right postcode — you are not competing in the two-hundred-thousand market. You are competing in a much smaller one with the same employers and better terms.
What would prove this wrong
The forecast rests on four assumptions. Each has a public signal attached, and we will revise this page against them rather than quietly restate the conclusion:
- Eurostat ICT-specialist growth falls below two percent. The net-creation figure would move toward one hundred thousand and the market would tighten sharply for juniors.
- Compliance deadlines slip materially. Delay or dilution of AI Act, NIS2 or CSRD obligations would remove the most reliable component of demand.
- Chips Act fab schedules slip again. Semiconductor hiring is a construction-schedule derivative; a two-year slip moves thousands of roles out of the window.
- Assisted engineering raises output per engineer faster than demand grows. The measurable version of this is headcount flat with delivery volume rising — watch reported output per employee, not vendor claims.
None of those four is speculative in the sense of being unobservable. They are all published, dated and checkable, which is the only standard on which a forecast deserves to be believed at all.