Generative-AI users
602mState Council reporting CNNIC · Dec 2025 ↗
people · December 2025 survey point
Innovation · industry, adoption and governance
Artificial intelligence industry and adoption
China's AI expansion is visible in industry scale, computing capacity, public adoption, regulatory filings and patent activity—but each metric measures a different part of the ecosystem.
User penetration
42.8%
CNNIC release wording · December 2025
AI core-industry scale
>1.2MIIT · 2025 industry snapshot ↗
trillion yuan · reported lower bound, 2025
AI enterprises
>6,200MIIT · 2025 industry snapshot ↗
enterprises · source-defined scope · 2025
Intelligent-compute capacity
1,590State Council · intelligent compute ↗
EFLOPS · 2025
Page method
Five statistical objects, five different meanings
Industry scale is measured in yuan, compute in EFLOPS, users through surveys, filings through administrative records and patent families through an international landscape study. They share a topic, not a common unit or objective weight.
No composite index
No normalization-and-sum score
Industry scale is not GDP contribution
Compute capacity is not model performance
Public adoption
Generative-AI users rose rapidly across three survey points
The chart retains only published CNNIC survey points—no monthly interpolation. The left axis shows users and the right axis the source-reported penetration rate.
Source: Xinhua reporting CNNIC · Jun 2024 ↗ · State Council reporting CNNIC · Jun 2025 ↗ · State Council reporting CNNIC · Dec 2025 ↗ · three actual survey points; no interpolationchinainfigures.com
Administrative records
Filed services and registered apps cannot be added
Each group separates additions during 2025 from the cumulative total at year end. The two groups describe different objects.
| Statistical object | New in 2025 | Cumulative at year end |
|---|---|---|
| Generative-AI service filings | 446 | 748 |
| Registrations of applications or functions calling filed models | 330 | 435 |
Source: CAC · 2025 announcement ↗ · new is an annual flow; cumulative is a year-end stockchinainfigures.com
International innovation comparison
Generative-AI simple patent families, 2014–2023
WIPO attributes cumulative simple patent families by inventor location. Volume reflects patent activity, not model quality or commercial value.
Source: WIPO · GenAI patent landscape ↗ · cumulative simple patent families · inventor locationchinainfigures.com
2025 industry and infrastructure snapshot
Three snapshots, not a fabricated time series
| Metric | 2025 value | What it indicates | What it does not indicate |
|---|---|---|---|
| AI core-industry scale | >1.2 trillion yuan | Scale of source-identified industry activity | GDP contribution or all AI revenue |
| AI enterprise count | >6,200 | Source-identified enterprise coverage | Pure-play firms, active models or profitable firms |
| Intelligent-compute capacity | 1,590 EFLOPS | Official intelligent-compute infrastructure scale | Utilization, efficiency, training cost or model performance |
Exact survey observations
Generative-AI adoption
| Survey date | Users | Penetration | Denominator note |
|---|---|---|---|
| Jun 2024 | 230 million | 16.4% | share of total population in the source wording |
| Jun 2025 | 515 million | 36.5% | penetration rate as reported by CNNIC |
| Dec 2025 | 602 million | 42.8% | penetration rate as reported by CNNIC |
Exact international comparison
WIPO generative-AI simple patent families
| Rank | Inventor location | Simple patent families |
|---|---|---|
| 1 | China | 38,210 |
| 2 | United States | 6,276 |
| 3 | Republic of Korea | 4,155 |
| 4 | Japan | 3,409 |
| 5 | India | 1,350 |
Industry scope
Scale is not value added
The source reports a lower-bound core-industry scale. This page does not present it as AI contribution to GDP or calculate precise growth from it.
Compute scope
EFLOPS is not a capability ranking
Compute capacity does not show actual utilization, chip efficiency, training cost, network/storage constraints or model performance.
User scope
Users are not daily active users
Survey users do not reveal frequency, payment, productivity or how many products one person uses.
Filing scope
Services, models and apps are different
Filed services and registered applications/functions are different objects. They cannot be added and do not equal active underlying models.
Patent scope
Volume is not quality
Simple-patent-family volume reflects filing and protection activity, not model performance, commercial value, citation impact or deployment.
Downloads and revisions
Use the data and track changes
The page CSV retains years, units, source IDs, verification status and comparability warnings. Each chart also offers its current data as CSV, PNG or SVG.
Page data
Download full CSV
Includes registered formal points, available historical series and scalar derived metrics. UTF-8 encoded for spreadsheets and analysis tools.
Download CSV ↓Data file updated: 5 Aug 2026
Citation
Copy a traceable citation
The title, latest formal data year and source organizations come from existing publication metadata; the access date and current page URL are added when you click.
Copy uses the URL currently open in your browser.
Public record
Revision and update timeline
Artificial-intelligence industry and adoption dataset prepared
Prepared the AI industry and computing snapshot, three CNNIC adoption observations, 2025 filing and registration counts, and WIPO's 2014-2023 patent-family comparison for a bilingual indicator page.
Related sources: Ministry of Industry and Information Technology of China ↗ · The State Council of the People's Republic of China ↗ · Xinhua News Agency, reporting CNNIC ↗ · The State Council of the People's Republic of China, reporting CNNIC ↗ · The State Council of the People's Republic of China, reporting CNNIC ↗ · Cyberspace Administration of China ↗ · World Intellectual Property Organization ↗
Sources and definitions registered; automated checks passed
Data downloads and revision timeline added
Added page-level formal-data CSV files, chart-level CSV/PNG/SVG exports and a bilingual revision timeline to all indicator pages.
Sources and definitions registered; automated checks passed