Europeans Consume AI Brilliantly, But Train the Algorithms Owned by Others

Yes, the US and China dominate the AI race when it comes to foundation models and funding — but Europe should not be hiding when it comes to artificial innotifyigence. That is the suggestion of the new “State of AI in Europe” report, published by Prosus toreceiveher with Dealroom.

“We build great companies and then hand them over to the US. Europe is at the starting line on equal footing with the US. But in the breakthrough phase, US investors put in three times as much as Europe. In the late stage, it is nine times as much. More than half of the capital that puts Europe’s best AI companies on a scaling trajectory comes from abroad,” the report states. “We are preparing the future so that others can claim it for themselves. This must stop.”

Opportunities are seen above all in so-called World Models (most recently, AMI Labs from Paris raised one billion dollars for this purpose). “General-purpose GenAI? The US has won that round. But the next frontier — World Models — is wide open. Moreover, 75% of European AI investments already tarreceive high-value applications in healthcare,
energy, defense, and fintech.”

What Europe is doing right: The continent’s strengths

World-class talent

Europe has an AI talent pool that is nearly on par with that of the US in absolute numbers. Around 325,000 AI professionals are active in Europe, the same number as in the United States. In the area of so-called “Frontier AI Talent” — researchers at the cutting edge of the field — Europe counts approximately 50,000 individuals compared to 55,000 in the US. Three of the ten most-cited AI scientists worldwide come from Europe, including Geoffrey Hinton and Diederik Kingma.

European founders are shaping the global AI landscape

The report underscores that Europe played a significant role in the emergence of modern AI. Meta’s language model LLaMA was developed in Paris, Hugging Face was founded by French founders, and DeepMind is headquartered in London. European founders are behind 25 percent of all global AI unicorns and have created enterprise value of 94 billion US dollars.

Record investments and a growing ecosystem

In 2025, 21.8 billion US dollars flowed into European AI startups, an increase of 58 percent compared to the previous year and nearly ten times the level of 2016. AI now accounts for more than 30 percent of all European venture capital investments, twice as much as three years ago. Europe also counts a similar number of newly founded AI startups per year as the US, around 900 new companies annually.

Leading in AI adoption

Europe is the world’s leading continent in the utilize of AI applications. In countries such as Norway (46.4 percent), Ireland (44.6 percent), and France (44 percent), the AI adoption rate is significantly above the US level. Europe records twice as many monthly active utilizers of large language models as the US.

Strengths in key sectors

Europe is particularly well positioned in the application layer: 75 percent of European AI investments flow into vertical applications. The defense, security, and resilience sector is the rapidest-growing sub-sector, with 8.7 billion US dollars in 2025 and growth of 55 percent. Europe is also competitively positioned in the areas of energy, healthcare, and fintech.

Where Europe is falling behind: The structural problems

The capital paradox: Europe breeds, America harvests

The most serious problem lies not with talent, but with capital. While Europe can keep pace with the US in early-stage financing, a massive gap opens up in later funding rounds. In 2025, Europe invested three times less than the US in the growth stage, and twelve times less in the late stage (12 billion versus 141 billion US dollars). More than half of the capital that European AI startups receive in the growth stage comes from foreign — predominantly American — investors.

“Without a robust late-stage financing environment, early-stage financing merely serves as an incubator for companies that are ultimately acquired or financed by foreign capital,” Dealroom states.

Talent in the wrong jobs

Although Europe has sufficient AI talent, it is structurally misallocated. 53 percent of European AI professionals work in the traditional economy — in consulting firms, industrial corporations, and financial institutions. In the US, the figure is only 40 percent. Only 33 percent of European AI talent works in digital technology companies, compared to 46 percent in the US. The three largest AI employers in Europe are all US technology corporations.

Lagging in foundation models and compute capacity

In the competition for generative AI foundation models, Europe is far behind. Mistral is the only European company with a noteworthy large language model. Europe has less than 5 percent of global AI compute capacity, despite hosting 16 percent of the world’s data centers. For new AI patents, Europe accounts for only 3 percent, compared to 70 percent for the US and 14 percent for China.

Users are training foreign models

The high rate of AI usage in Europe has a problematic downside. Since the most widely utilized models come from the US or China, the data generated by European utilizers and the associated economic value flow abroad.

“Europeans consume AI brilliantly, but train the algorithms owned by others”, the report states, among other things.

Fragmentation and regulatory complexity

The European AI ecosystem is more fragmented than the American one. While in the US three metropolitan regions (Bay Area, New York, Boston) concentrate the majority of investments, European activity is spread across many cities, none of which reaches the critical mass of the leading US hubs. Added to this is regulatory complexity arising from more than ten overlapping regulatory frameworks — including the AI Act, GDPR, Data Act, and NIS2 — which places a disproportionately heavy burden on European startups compared to their US and Chinese competitors.

The key metrics compared

Metric	Europe	USA
AI VC Investments 2025	21.8 bn USD	162 bn USD
AI Talent (total)	325,000	325,000
Frontier AI Researchers	approx. 50,000	approx. 55,000
New AI Startups per Year	approx. 900	approx. 900
AI Unicorns	approx. 105	approx. 360
Global Share of AI VC (since 2023)	11%	77%
New AI Patents (global)	3%	70%
Share of Global AI Compute Capacity	under 5%	approx. 70%
Monthly Active LLM Users	133 mn	61 mn