Who actually builds AI in Europe

We mapped 27,444 engineers across 348 AI companies building in Europe. Only 14% ever worked at Big Tech. Here is where the talent really comes from, and how to hire against it.

By Laurence Sangarde-Brown · July 10, 2026

The Engineers of European AI

The prestige story says European AI is built by ex-Google and ex-DeepMind engineers. We reconstructed the workforce profile by profile, and the story does not survive contact with the data.

You are staffing an AI company in Europe and you write the job spec everyone writes: ex-FAANG preferred, elite degree, ideally a stint at a frontier lab. It feels like a filter for quality. It is really a filter for scarcity, and you are about to pay for it.

We took the Top-100 global VC cohort, pulled every AI company they have funded, kept the ones with a real European engineering team, then added Europe's most-watched startups. The result is 348 companies and 27,444 engineers, counted from employment history rather than surveyed. The composition looks nothing like the pitch-deck version.

The number that breaks the narrative

Only 14% of these engineers ever worked at classic Big Tech, and fewer than 1% came from a frontier lab. Elite-university pedigree sits at 25%, and just 8% hold a PhD. On every axis the market is less rarefied than the story you have been told.

14 in 100. The share of Europe's AI engineers who ever worked at classic Big Tech. Benchmark only against them and you are fishing in a seventh of the pond.

So where do they actually come from?

Follow the previous employer and an industrial map appears. IBM is the single largest feeder, ahead of Microsoft and Google, and right behind sit the consultancies and telecoms. Europe is not importing an AI workforce so much as retraining the large, experienced one it already had.

Widen the filter past the FAANG logo. The engineers you want more often carry a decade of shipping real systems than a two-year hyperscaler stint.

Mine the enterprise and consultancy base. Ericsson, Nokia, Accenture and Capgemini are quietly losing exactly the people you are trying to hire.

Treat pedigree as one weak signal, not a gate. Recent building history predicts output far better than a crest on a degree.

Big Tech is the largest single feeder, but a clear minority of the whole. Precision beats narrative.

The firms that keep chasing the ex-FAANG badge will keep overpaying into a thin, contested pool. The ones that map where capability actually sits, the subject of this series, will out-hire them on the same budget.

Composition is a data question. Treating it as one is the edge.

Every figure here comes from The Engineers of European AI, our study of 27,444 engineers across 348 AI companies building in Europe.