Compare the 2026 job market for an early-career AI engineer in Amsterdam, Rotterdam, Eindhoven and Paris — for a non-EU (Pakistani) Sorbonne CS master's graduate (Sept 2027) with ~4 years of industry experience (LLM agents, RAG, Rust, AWS, Kubernetes), English only. Requested per city: current English-only AI/ML/LLM openings for 0–3 years, junior pay median and P75, net monthly pay (Dutch cities also under the user's 27% ruling scenario for new rulings from 2027), one-bedroom rent, net after rent, top employers, IND sponsor status, 2025–2026 layoffs or freezes, and winners.
This report covers all four requested cities. It rests on a vacancy-level audit of 91 current English-only AI/ML/LLM openings for 0–3-year profiles (47 Paris, 19 Amsterdam, 14 Eindhoven, 11 Rotterdam) across 38 ranked employers, 124 compensation observations, 32 published one-bedroom rent estimates (eight per city; representative rent is the median of these published estimates, not an official transaction average), 2026 tax-calculator evidence, and seven 2025–2026 workforce-news items. Each employer/opening row is single-sourced from one aggregator or job board.
| City | Audited English-only openings (0–3 yr) | Junior annual median (EUR) | P75 (EUR) | Net/mo at median, ordinary | Net/mo, 27% ruling (2027) | 1-bed rent/mo | Net after rent, ordinary | Net after rent, ruling | Verdict |
|---|
Bars: audited English-only openings per city (counts) and estimated monthly net after representative rent (EUR). Net figures are estimates for a single employee with no children, 2026 tax rules, anchored to published 2026 calculators; ruling column applies the user's 27% scenario for rulings starting 2027. Dutch city-level junior-AI percentiles were too sparse to compute defensibly, so medians are labelled proxy roles and P75 is left as insufficient evidence.
Dutch highly skilled migrant hiring generally requires an IND recognised sponsor. "Sponsor confirmed" means confirmed in this audit's collected evidence; "not confirmed" means the audit found no confirmation, not a definitive non-sponsor. Paris does not use IND: French work-authorisation/talent-passport routes apply, so IND is marked N/A. A 2026 snapshot is not a guarantee of conditions at a September 2027 graduation.
Most audited English-only openings: Paris, with 47 of the 91 audited vacancies — more than the three Dutch cities combined.
Highest estimated monthly net after representative rent: Amsterdam in both scenarios — about €2,215 under ordinary 2026 taxation (Eindhoven ≈€1,935) and about €3,005 under the 27% ruling scenario (Eindhoven ≈€2,595). Paris trails at ≈€1,469: the lowest rent cannot offset the lowest junior median. This hinges on the Picnic ML-L3 proxy for Amsterdam; the underlying junior evidence is sparse and heterogeneous.
Method: vacancy-level audit of 91 current English-only AI/ML/LLM openings (0–3-yr scope) across 38 ranked employers in four cities, collected from job aggregators and boards in Sept–Oct 2026; 124 compensation observations (Levels.fyi, Glassdoor and similar) filtered to city-level junior/entry records — Dutch medians are proxy software/ML roles because exact junior-AI percentiles existed only for Paris; representative rent is the median of 32 published 2026 one-bedroom estimates (8 per city); net pay is an estimate for a single, childless employee under 2026 tax rules anchored to published calculators, excluding holiday-allowance timing, pension, equity taxation, health-insurance premiums and France's prélèvement timing; the 27% ruling column models the user's scenario for rulings starting 2027. Layoff timeline: 7 deduplicated 2025–2026 news rows. Broad job-board counters (e.g. 916 LinkedIn Amsterdam results) lack language filters and were excluded from the comparison. Evidence-text columns were cut for space.