Filters scope the adoption and sector section. Compute and research charts are global.
Task 1 core: AI adoption in strategic sectors
AI uptake by strategic sector
Share of enterprises (10+ employees) using at least one AI technology, NACE groups weighted by firm counts per the JRC note
Source: Eurostat isoc_eb_ain2, JRC weights from Annex 7. Sectors marked * are only partly covered by the survey.
Enterprise AI adoption over time
Share of enterprises (10+ employees) using at least one AI technology, all activities except finance
The list of AI technologies in the questionnaire changed between the 2021 and 2023 surveys, so the first step is partly a definition change. No survey in 2022.
Source: Eurostat isoc_eb_ai, indicator E_AI_TANY.
Why enterprises that considered AI did not adopt it
Share of enterprises that ever considered AI, by reported reason
Source: Eurostat isoc_eb_ai, unit PC_ENT_AI_EC. Multiple answers possible.
Which AI technologies enterprises use
Share of all enterprises (10+ employees), by technology type
Source: Eurostat isoc_eb_ai, unit PC_ENT.
Regional view: enterprise AI adoption by NUTS 2 region
Share of enterprises (10+ employees) using at least one AI technology. Regional reporting is optional, so most large member states publish a national figure only.
Source: Eurostat isoc_r_eb_ain2. Boundaries: Eurostat GISCO NUTS 2024 1:20M, © EuroGeographics. No tile server is used; the base map is a local GeoJSON.
Table view
Enablers: compute
Share of TOP500 performance by region
Sum of Rmax (Linpack) of listed systems, per edition, percent
Voluntary benchmark submissions. China's share moved from 1.4% to 12.7% in one edition because a single new system submitted a run.
Source: TOP500.org editions June 2022 to June 2026. Reuse terms to be confirmed with TOP500.org.
Largest EU27 systems in the latest TOP500
Source: TOP500.org. Rmax in PFlop/s.
Documented AI GPU clusters
Clusters with a known location, marker area proportional to H100-equivalents
Source: Epoch AI, Data on GPU clusters (CC BY 4.0). Chinese clusters are anonymised in the public file and carry no coordinates.
Recorded AI compute capacity by country
Share of H100-equivalents in documented clusters, top 10 countries plus EU27
Source: Epoch AI, Data on GPU clusters. Share of recorded capacity, not of global capacity.
Enablers: research and models
AI research output by author country
Works with primary subfield "Artificial Intelligence", count of author-country assignments per year
Source: OpenAlex (CC0), subfield 1702, types article, preprint, review. Whole counting: a work with authors in two countries counts once for each.
Notable AI models by developer country
Models meeting Epoch AI's notability criteria, by year of release, fractional count for multi-country models
Notability is assigned with a lag, so the two most recent years are incomplete.
Source: Epoch AI, Data on AI models (CC BY 4.0).
People: generative AI use
Individuals using generative AI tools for work
Share of individuals aged 16 to 74, 2025, by country
Source: Eurostat isoc_ai_iaiu, indicator I_IUAIWP. First survey year; no non-European country is covered.
Reports
Monthly monitoring reports
Sector monitoring reports and reports on cross-cutting challenges, written by the Observatory team. PDF, English.
- Loading reports…
Reports are published after review by the Commission services.