19.95%
2025 · enterprises with 10+ employed
The EU is a regional aggregate of 27 member states; its survey figure is not an unweighted average of country percentages.
2025 · enterprises with 10+ employed
Broader Epoch catalogue, including variants; not a complete model census.
All R&D, not AI-only investment.
Same source, year and developer-country attribution. The broader catalogue includes versions and sizes. Neither number is the total number of models developed here.
Loading source records…
Selected projects checked against institutions and developers. These may overlap with Epoch; they are not added to its count.
French developer registration; individual release documentation is checked separately
Attribution sourceDeveloper launch identifies Aleph Alpha as a German AI company and names both 7B variants
Attribution sourceAina identifies BSC development and MareNostrum 5 training; base, instruction and quantized variants are distinct
Attribution sourceMinistry documents development by Polish scientific institutions and language adaptation; this does not establish all variants as trained from scratch
Attribution sourceFirst-party checks reviewed 2026-10-25 · selected coverage
Annual notable-model entries · developer country · 2020–2025
EU is a regional aggregate; US and China are countries. International collaborations may appear in multiple countries. Counts are not additive market shares or a national AI-strength ranking.Epoch AI · CC BY 4.0 · Complete years through 2025 · incomplete 2026 excluded
Business adoption across Europe — and the divide between small and large firms
○ Small: 10–49 persons · ● Large: 250+ · percentage of enterprises within each size class. Each point has its own denominator.
Percentage of enterprises using at least one AI technology. Selected non-financial business activities (NACE Rev. 2 C–J, L–N and 95.1). Enterprises with at least 10 persons employed. Not intensity of implementation, productivity impact or share of workers using AI.
Coverage is European survey participants, not a global adoption ranking. Small firms with fewer than 10 persons employed are excluded. Size classes and the total are separate survey populations; percentages must not be added. EU is an aggregate. Non-response, survey design and sector mix can affect comparisons.
Source flags are retained: b = break in time series; e = estimated; p = provisional; u = low reliability. Missing and confidential values are not replaced by zero. No 2022 observation is inferred.
Eurostat · isoc_eb_ai · source update 2026-06-15T11:00:00+0200 · captured 2026-10-23 · Eurostat reuse policy; attribution and stated exceptions apply
eurostat-enterprise-ai-size-v1 · SHA-256 f5f6d136c4ed04923d17a63c4b0b23d6999e7a4774450709ae214c24b7674357
Public and private investment target, including a €20bn gigafactory fund.
These indicators describe all research and development, not AI-only expenditure or AI’s measured economic impact.
2.26% of GDP · all R&D · 2023 · European Union
Comparison uses 2023, the latest common year for this selection. The latest national observation above may be newer.
World Bank WDI / UNESCO Institute for Statistics · % of GDP · all R&D4,741.41per million people · all R&D · 2023 · European Union
Comparison uses 2022, the latest common year for this selection. The latest national observation above may be newer.
World Bank WDI / UNESCO Institute for Statistics · per million people · all R&DEpoch AI · CC BY 4.0 · Model comparisons stop at 2025; incomplete 2026 is excluded. No forward-filled research observations.
Epoch’s notable-model selection · developer country · international projects can appear in multiple countries
Mistral OCRMistral AIPixtral LargeMistral AIMistral Large 2Mistral AIDeepL LLMDeepLMathstralMistral AIMistral LargeMistral AIMixtral 8x7BMistral AISeamlessM4TFacebook,INRIA,University of California (UC) Berkeley