How many European businesses actually use AI?
19.95% of EU enterprises used at least one AI technology in 2025 — up from 13.5% a year earlier. But adoption splits three-fold by size: 17% of small firms vs 55% of large ones.
Eurostat enterprise survey 2025 (enterprises with 10+ employees), US Census BTOS Nov 2025 – Jan 2026 · verified as of
Adoption by country (% of enterprises, 2025)
The north–south-east gradient is stark: Denmark adopts at five times Bulgaria's rate. For B2B AI founders this is a market-entry sequencing map.
| Metric | Value | Segment | Period | Source |
|---|---|---|---|---|
| Denmark | 42.0% | EU leader | 2025 | ec.europa.eu |
| Belgium | 34.5% | 2025 | ec.europa.eu | |
| Austria | 30.0% | 2025 | ec.europa.eu | |
| Germany | 26.0% | 2025 | ec.europa.eu | |
| Estonia | 23.4% | CEE leader | 2025 | ec.europa.eu |
| EU-27 average | 19.95% | up from 13.5% in 2024 | 2025 | ec.europa.eu |
| Czechia | 17.6% | was 4.6% in 2021 | 2025 | ec.europa.eu |
| Greece | 8.9% | 2025 | ec.europa.eu | |
| Bulgaria | 8.5% | EU floor | 2025 | ec.europa.eu |
The gaps that define the market
Small European businesses are the underserved segment: they adopt at a third of the large-enterprise rate, and when they do adopt, they go far less deep.
| Metric | Value | Segment | Period | Source |
|---|---|---|---|---|
| Adoption gap by size | 17.0% vs 55.0% | small (10–49) vs large (250+) enterprises | 2025 | ec.europa.eu |
| Depth gap: using 3+ AI technologies | 6.5% vs 33.6% | small vs large enterprises | 2025 | ec.europa.eu |
| Considered AI but did not adopt | 14.2% | of EU non-adopter enterprises | 2025 | ec.europa.eu |
| Lowest-adoption sector: construction | 10.8% | vs real estate 24.8% | 2025 | ec.europa.eu |
Why considerers walk away (EU-27, 2025)
The top barrier is not cost — it is legal uncertainty. A compliance-first pitch beats a price-first pitch in this market.
| Metric | Value | Segment | Period | Source |
|---|---|---|---|---|
| Lack of legal clarity | 53.6% | of enterprises that considered AI | 2025 | ec.europa.eu |
| Data availability or quality | 43.5% | of enterprises that considered AI | 2025 | ec.europa.eu |
| Cost | 38.4% | of enterprises that considered AI | 2025 | ec.europa.eu |
The US mirror (Census BTOS, Nov 2025 – Jan 2026)
The feared job apocalypse is not in the microdata: among AI-using US firms, augmentation outnumbers pure substitution nine to one.
| Metric | Value | Segment | Period | Source |
|---|---|---|---|---|
| US firms using AI in a business function | 18% | all firm sizes | Nov 2025 – Jan 2026 | www2.census.gov |
| US firms where workers use AI in tasks | 23% | Nov 2025 – Jan 2026 | www2.census.gov | |
| Large US firms (250+) using AI | 37% | 2026-05 | census.gov | |
| AI-using firms reporting AI-related employment decrease | 2.0% | Nov 2025 – Jan 2026 | www2.census.gov | |
| AI augments worker tasks vs replaces them | 44% vs 5% | of AI-using firms | Nov 2025 – Jan 2026 | www2.census.gov |
Why "AI adoption" ranges from 8% to 70%+ in headlines
Three different questions hide behind one phrase. Enterprise surveys (Eurostat: 19.95%, Census: 18%) ask whether the firm uses AI in business processes — with different size thresholds, so even these two aren't directly comparable. Workforce surveys (41% of the US workforce reports work-related generative-AI use) count people, not firms. And function surveys (70% of organizations use genAI in at least one function, Stanford HAI) count any team anywhere in large organizations. All three are legitimate; quoting one as if it were another is how most AI-market slides mislead.
Eurostat — AI technologies in enterprises · US Census — BTOS AI supplement
Methodology
Every number on this page is a fact node in the God of Startups Atlas — a knowledge graph where each metric carries its source URL, reporting period and evidence tier. Tier 1 here means official statistics: Eurostat's enterprise ICT survey and the US Census Bureau's Business Trends and Outlook Survey. Survey scope differences are stated, not smoothed over.
Related benchmarks
Where does your model sit on these curves?
Map your startup’s numbers against this data inside God of Startups. Your idea is checked against the market, never shared with it.