We reconstructed 27,444 engineers across Europe's AI companies, profile by profile, from the biggest labs to the most-watched startups. Here is where they actually came from.
0AI companies in cohort
0Europe-based engineers indexed
0ever worked at Big Tech
01 The headline
The narrative says Europe's AI runs on ex-Google defectors. The data says 14%.
Ask a founder where Europe's AI engineers come from, and you will hear the same answer every time: poached from Google, DeepMind, Meta. It is a good story. It is also mostly wrong. We ran the numbers against a named set of 21 hyperscaler entities instead of taking the story on faith, and only 14% of the pool previously worked at classic Big Tech. If you are building, hiring or investing in European AI, that gap between the story and the data should change where you point your sourcing budget.
Key insights
Only 14% of Europe's AI engineers previously worked at classic Big Tech, not the ex-Google story most founders assume.
The analysis spans 348 companies and 27,444 engineers building AI across Europe, from the biggest labs to the most-watched startups.
Enterprise, telco and consultancy incumbents feed the pipeline, led by IBM (530), ahead of Microsoft and Google.
Big Tech pedigree concentrates in a handful of frontier labs and spinouts: Poolside runs hottest at 44%, most others sit in single digits.
Polytechnic Bucharest ranks third among feeder universities, with Central and Eastern European technical schools supplying real volume.
Team pedigree peaks at Series A, where top-100 university share jumps to 51%, then dilutes through later stages.
Elite-university and PhD share among Europe's AI engineers.
02 The size of the pool
27,444 engineers across 348 companies building AI in Europe.
Everyone has an opinion on how deep Europe's AI talent pool runs. Almost nobody has actually counted it. So we did: start from the top-100 global VC cohort and funnel down to 273 companies with a real European engineering team, then add Europe's most-watched AI startups on top. The full analysed set is 348 companies and 27,444 engineers. That is the denominator for every claim in this report.
Start with the top-100 global VC cohort and pull every company they have funded.
Keep the ones carrying an AI signal.
Keep only those running a genuine European engineering team of at least ten Europe-based engineers.
The top-VC core as a funnel: funded companies, AI-signal companies, then those with a real European engineering team. Europe's most-watched AI startups are added on top to reach the full set of 348.
03 Where they come from
An enterprise, telco and consultancy pipeline, not a Big Tech exodus.
Rank the top feeder organisations by the number of distinct Europe-based engineers who previously held a role there, and the hyperscalers do not even lead the table. IBM leads with 574, Microsoft 528, Google 498, then Accenture, Amazon and Deloitte.
The pipeline is dominated by enterprise, deep-tech and industrial incumbents and by the global consultancies.
This is Europe retraining its own industrial and IT-services base, not a Silicon Valley talent exodus.
Top feeder organisations by distinct Europe-based engineers previously employed there.
IBM alone has fed more engineers into Europe's AI companies than Google. Build a sourcing plan around ex-hyperscaler talent, and you are fishing in the wrong pond.
04 Where Big Tech does cluster
Concentrated in the frontier labs, thin everywhere else.
The Big-Tech-heavy teams do exist, but they cluster hard. Among companies with at least forty Europe-based engineers, Poolside runs hottest at 44%, followed by Cohere, Anthropic and Eleven Labs. Most of the broad cohort sits in single digits. Treat these as outliers, not the pattern, and plan your pitch accordingly.
Share of Europe-based engineers who came from classic Big Tech. Companies with 40+ Europe-based engineers.
Outside a handful of frontier labs and hyperscaler spinouts, a Big-Tech-heavy pitch to candidates will not land. 86% of the cohort has never worked at a hyperscaler.
05 Where they studied
Oxbridge and Imperial still top the table, but the supply spans all of Europe.
Cambridge, Imperial, Oxford, UCL, TUM and ETH lead, exactly as the prestige story predicts. Fine, that part checks out. But the broad universe surfaces a second tier the elite reports miss entirely: Polytechnic Bucharest ranks third by headcount, and the Central and Eastern European technical universities (Budapest, AGH Kraków, Bucharest) sit alongside Aalto as real sources of the region's engineering volume, not just a discount option.
Top 20 universities by Europe-based engineers. Emerald marks the emergent non-elite (outside global top-100) second tier.
06 Four ways to build a team
Is Anthropic's European team built like Databricks'? No, and the difference is measurable.
Plot every company with at least forty European engineers on two axes, the share from Big Tech and the share holding a PhD, and European AI teams sort themselves into four distinct archetypes. Know which one you are, or which one you are competing against, and your recruiting playbook changes with it. Bubble size is engineer count.
Each company with 40+ Europe-based engineers, plotted on industry pedigree (x) and research depth (y). Dashed lines mark the archetype thresholds.
07 Where they sit
The UK leads; Eastern Europe is the volume engine.
Group engineers by where they physically work, not by company HQ, and the real geography snaps into focus. The UK hosts 7,934 engineers, 29% of the cohort, and leads on pedigree too. Switzerland is tiny but the most elite (48% top-100, 20% PhD). Poland, Romania and Hungary emerge as large-volume delivery hubs with low formal-elite pedigree, the kind of market that never makes the trade press.
Europe-based engineers by country of work. Dot area scales with engineer count; the ranked list below gives each market's headcount and top-100 university share.
Poland, Romania and Hungary together host nearly 3,000 engineers at a fraction of Switzerland's pedigree cost. The volume play sits east, not in the capitals with the best press coverage.
Guest perspective · inside the Polish AI scene
"Don't score Polish talent on delivery or polish. Score it on substance, or you will systematically underrate strong teams and lose them to whoever reads them correctly."
The data shows Poland as one of Europe's high-volume engineering hubs. To ground that from the inside, we asked Karol Lasota of Inovo.vc, one of Poland's most active early-stage investors, how the scene actually looks on the ground.
Q1 Where is Poland's AI talent concentrated, and what's driving it?
It is mostly Warsaw. There are strong teams in Wroclaw and Krakow too, but the biggest concentration of AI talent, both Polish and international, is gathering in Warsaw. Many of the key research centres are there, with new ones appearing, and it is heavily network-effects driven: every hot opportunity brings another, and one success brings the next.
Q2 Which Polish teams stand out on engineering quality, and how are they built?
ElevenLabs is the clearest proof point; their success speaks directly to the quality of their research and engineering teams. Among the still-early teams, Viktor is a good example. What stands out is who they hire: people with unique personalities who pursue greatness and have often already tasted it. The ambition to be part of something generational, and real passion for it, not just a good job.
Q3 Poland is known for strong technical universities and competitive salaries. Does the cost story still hold?
Looking at Poland purely through the lens of salary cost is a lazy read, and it does not hold up anymore. You are competing for AI talent against the entire world now, not just regionally. What Poland has is momentum: fast growth, good opportunities, safety and relatively low tax rates. More and more top companies are offering real opportunities here, and that combination is bringing top people back to Poland, or pulling them in for the first time.
Q4 Are the best Polish AI engineers staying local, moving to European hubs, or being pulled toward US remote roles?
Many young entrepreneurs are moving to San Francisco, and I think that is a good thing; they get exposure to state-of-the-art work and a level of velocity and ambition that is hard to find elsewhere. For talent more broadly, more interesting opportunities are appearing in Poland, so fewer people are leaving than before. If anything, more are coming back.
Q5 What would you tell a European VC or scale-up about accessing Polish talent that isn't obvious from the outside?
Eastern European founders and engineers are culturally less salesy: less prone to performative confidence, less inclined to oversell. To a US or Western European investor calibrated on American-style self-presentation, that can read as weaker conviction or lower ambition. It is not. The practical takeaway: do not score Polish talent on delivery, score it on substance directly, or you will systematically underrate strong teams and lose them to someone who reads them correctly.
About the contributor
Seed-stage venture · Central & Eastern Europe
Inovo is a Warsaw-based seed-stage venture firm and one of Central and Eastern Europe's most active early-stage investors, now deploying a €107m third fund across the region. Its core areas of conviction include AI and machine learning, developer tools and data science, the last of which Karol leads, and the firm publishes some of the most-cited research on the CEE tech ecosystem, including its State of DevTools in CEE. That gives Inovo an unusually close read on where the region's AI engineering talent is forming, and who is building with it.
TechTree Research
Get the next report before it goes public.
One email per report. Composition, migration and hiring intelligence from the TechTree knowledge graph, with the data shown, not summarized.
No spam, unsubscribe any time. Read the archive at techtree.dev.
Done. You are on the list.
08 Who backs the best
Balderton and Lightspeed back the highest-pedigree teams.
Because this universe is defined by investors, we can compare them head to head, and the gaps are wide. Balderton (33% top-100) and Lightspeed (32%) back the highest-pedigree European teams, roughly double Accel's rate. Sequoia has both the largest base (5,225 engineers) and the deepest research talent (11% PhD). Growth investors, by contrast, back the most engineers at the lowest pedigree, which is exactly what you would expect at that stage.
Investors ranked by top-100 university share of their European AI portfolio. Dot area scales with engineer count.
09 By stage of company
A European AI team is at its most elite at Series A. Then it dilutes.
Bucket Europe's most-watched AI startups by the highest priced round they have reached, and a clean arc emerges. Pedigree and research depth peak at Series A and then dilute. If you are hiring at Series B or beyond, that dilution is not a red flag, it is the default trajectory of scaling.
Teams scale roughly tenfold from Seed to Series C+.
Top-100 university share jumps to 51% at Series A before settling.
PhD share peaks at 21% and more than halves by Series C+.
Average team size (bars) against top-100 university and PhD share (lines) across funding stage.
If your pedigree bar is set at the Series A level, hold it there deliberately. Every company loosens it by default once the team passes 150 people, and that is a choice, not an accident.
10 Europe's most-watched AI startups
The teams behind the names, and how they differ by the type of AI they build.
Zoom in on the startups everyone is watching, from Mistral and ElevenLabs to Helsing and Wayve, and the engineering picture sharpens fast. 8,709 engineers build these companies. The previous section cut them by funding stage. Here we cut them by type of AI, and the composition shifts markedly from one category to the next, in ways that should reset how you benchmark each one.
8,709engineers in the most-watched startups
36.5%hold a top-100 university degree
11.8%hold a PhD
Engineers by type of AI the company builds. Applied and vertical products employ the most; deep-science teams are smaller but the most elite.
Applied & vertical
36 companies · 3,805 engineers
31% top-100 university · 7% PhD · 11% ex-Big Tech
Foundation & infra
27 companies · 2,454 engineers
33% top-100 university · 14% PhD · 18% ex-Big Tech
Robotics & physical
5 companies · 1,261 engineers
30% top-100 university · 12% PhD · 13% ex-Big Tech
Science & health
15 companies · 599 engineers
51% top-100 university · 21% PhD · 13% ex-Big Tech
Defence & security
3 companies · 594 engineers
42% top-100 university · 12% PhD · 13% ex-Big Tech
Type predicts pedigree more than fame does. Science and health teams run at 51% top-100 and 21% PhD, foundation and infrastructure teams pull hardest from Big Tech (18%), and the large applied and vertical bucket is the least research-heavy (7% PhD). Match your benchmark to the category, not the headline.
The largest engineering teams among the most-watched startups, by current engineer count.
Big-Tech magnets
Highest ex-hyperscaler share
The frontier model teams poach hardest: Kyutai (45%), Prior Labs (43%), Poolside and Tessl (42%), Cradle (41%).
Research-dense
Highest PhD concentration
Deep-tech runs on doctorates: Materials Nexus (75%), Boltzbit (50%), Emmi AI (47%), inait and Kyutai (44-45%).
Global from day one
Teams are not only European
The most-watched startups already run large offshore engineering: 789 engineers in the US, 128 in India, 101 in Tunisia (InstaDeep's roots).
The UK core
Elite and concentrated
The UK holds 2,483 of these engineers at 51% top-100 and 15% PhD, the densest elite cluster in the set; Switzerland is smaller but even more elite.
11 The company explorer
All 348 companies, every metric, sortable and filterable.
Every figure in this report decomposes to a company-level row, so you do not have to take our word for the rollups. Search by name or country, click any column to sort. Composition is computed on Europe-based engineers only, regardless of where the company is headquartered.
Company
HQ
EU eng.
% Big Tech
% Top-100
% PhD
Archetype
12 Strategic recommendations
What this means if you are building, hiring or investing in European AI.
The data points to three different playbooks depending on where you sit. None of them starts with a Big Tech pedigree filter, and that is precisely the point.
Strategic planning assumptions
Through 2028, enterprise, telco and consultancy incumbents will remain the largest single source of engineers for Europe's AI companies, ahead of the hyperscalers.High confidence
The ex-Big Tech share of Europe's AI engineering workforce will stay below 20% through 2028, even as frontier labs expand their European offices.High confidence
Team pedigree will keep peaking at Series A: growth-stage European AI companies that hold a Series A-level hiring bar past 150 engineers will remain rare exceptions.Moderate confidence
For founders
Recruit from IBM, Microsoft, Accenture and Nokia before chasing ex-Google candidates. They supply more of the pool and cost less to win.
Set your pedigree bar at Series A levels (51% top-100, 21% PhD) deliberately if research depth matters to your product. Left alone, it dilutes by default after that stage.
Build delivery capacity in Poland, Romania or Hungary rather than only the UK, where the pool is deep but the pedigree premium is smallest.
Benchmark your team against the archetype closest to your product, not against the frontier labs' 40%-plus Big Tech share, which is an outlier even inside that group.
For investors
Read team composition against Balderton and Lightspeed's 32 to 33% top-100 benchmark when assessing pedigree at term sheet stage.
Treat a heavy ex-Big-Tech team as a signal to investigate, not a quality signal by default. Only 14% of the broader pool carries that background.
Expect dilution of pedigree and PhD share past Series A, and underwrite growth-stage deals on execution and distribution, not on the team's academic pedigree.
Weight Central and Eastern European university pipelines (Aalto, Budapest, Bucharest Polytechnic) as a real, underpriced source of engineering supply, not a discount signal.
For recruiters and talent leaders
Build sourcing lists around enterprise, telco and consultancy alumni, the pipeline that actually supplies 86% of the pool.
Use the four team archetypes to write role-specific outreach instead of one generic "AI engineer" pitch across every company type.
Expand searches into Aalto, Budapest and AGH Kraków alongside the usual Oxbridge and Imperial shortlist to widen a thin candidate pool.
Calibrate pedigree expectations to company stage. A Series C+ team will not match Series A pedigree benchmarks, and holding it to that bar will slow every search.
Run this for your market
The same engine can map your role, your portfolio or your country.
Every chart in this report was computed from the TechTree knowledge graph on demand. Leave a work email and we will show you what it says about the market you hire in.
We reply personally to every request. No sequence, no SDR.
Done. We will be in touch shortly.
13 Method & confidence
Transparent, repeatable, decomposable to the company table.
This is one analysis of the engineers building AI in Europe, drawn from the TechTree workforce knowledge graph. Alongside the full landscape we profile Europe's most-watched AI startups and cut their engineering teams two ways: by the funding stage a company has reached, and by the type of AI it builds. For the most-watched startups, composition is computed on each company's full current engineering team.
An "engineer" is someone whose current role classifies as Computer and Mathematical or Architecture and Engineering, employed in a European country; 90% carry education data. "Big Tech" is a fixed set of 21 hyperscaler entities, so the 14% figure is measured against a named list, not a vibe, which is the whole point of the exercise.
Employment history, education, geography, tenure and funding are Tier 1 confidence, and we use them aggressively. Role and seniority are Tier 2, and we use them carefully. Investor attribution follows recorded funding history, so a single early cheque can attach a now-large company to a firm. Read the investor rankings as portfolio-shape signals, not league tables. Data as of 24 June 2026, drawn from the TechTree workforce knowledge graph.
Free · Full report
Read the full Engineering Composition of Europe's Top-VC-Backed AI Companies
27,444 engineers across 348 top-VC-backed AI companies with European engineering teams — origins, pedigree, archetypes, ecosystems and investor cohorts. Enter your work email to unlock the full report.
