AI Adoption by Country 2026
By Axis Intelligence Research
Co-author: Sarah Mitchell | Last updated: July 27, 2026 | License: CC BY 4.0
The UAE leads all 147 tracked economies with 64.0% of its working-age population using generative AI — more than double the 28.3% recorded in the United States, which ranks 24th despite hosting more AI data centers than any other country on Earth. According to Axis Intelligence Research’s analysis of Microsoft, OECD, Eurostat, and Stanford HAI data, global AI usage reached 16.3% of the world population in H2 2025, but the gap between the highest and lowest adopters spans 59 percentage points, and it is widening, not narrowing.
Quick Answer: Which Countries Lead AI Adoption in 2026?
At the population level (share of working-age adults using generative AI tools at least once in a period), the UAE leads at 64.0%, followed by Singapore at 60.9% and Norway at 46.4%, per Microsoft’s AI Diffusion Report H2 2025. At the enterprise level, Denmark leads OECD economies at 42.0% of firms using at least one AI technology, per Eurostat’s December 2025 release. The global population-level average stands at 16.3%. According to Axis Intelligence Research’s analysis across all four major measurement frameworks — Microsoft telemetry, OECD firm surveys, Eurostat enterprise data, and McKinsey’s executive survey — no single country leads uniformly across all dimensions: the countries with the highest population-level usage are not the same as those with the highest enterprise deployment rates, and neither group overlaps cleanly with the leaders in AI investment.
Key Findings
- According to Axis Intelligence Research’s cross-source analysis of Microsoft’s H2 2025 AI Diffusion Report, the UAE reached 64.0% working-age population AI usage in H2 2025 — more than double the United States’ 28.3%, a gap of 35.7 percentage points that widens when normalized to each country’s baseline.
- According to Eurostat’s December 2025 release of the ICT Enterprise Survey (dataset ISOC_EB_AI), 20.0% of EU enterprises with 10 or more employees used AI technologies in 2025, up 6.5 percentage points from 13.5% in 2024 — the sharpest one-year jump since EU measurement began.
- According to Axis Intelligence Research’s calculation combining Microsoft H2 2025 telemetry data and Eurostat enterprise data, Denmark exhibits what Axis Intelligence Research terms a negative Adoption Disparity Score: its enterprise AI adoption rate (42.0%) exceeds its population-level usage rate (28.7%) by 13.3 percentage points — a structural pattern shared by Finland (+10.5 pp) and Sweden (+1.7 pp) but reversed in the Netherlands (-5.7 pp) and Belgium (-1.5 pp).
- According to the OECD’s January 2026 ICT Access and Usage Database release, firm-level AI adoption across OECD countries reached 20.2% in 2025, up from 8.7% in 2023 — meaning adoption more than doubled in two years, with the ICT sector at 57.3% and professional services at 36.8%.
- According to Stanford HAI’s 2026 AI Index Report, the Global North’s working-age AI usage rate grew 1.8 times faster than the Global South’s in H2 2025 — the North gained +1.8 percentage points against the South’s +1.0 — widening an already-significant divide from 9.8 to 10.6 percentage points.
How Do You Measure AI Adoption by Country?
This is the question every analyst answering the headline misses. There are at least four distinct measurement frameworks in active use, each measuring something different, and their outputs are not directly comparable.
Microsoft’s AI Diffusion metric measures the share of working-age adults who used a generative AI product at least once during a reporting period, derived from aggregated, anonymized telemetry across Microsoft services and adjusted for OS/device-market share, internet penetration, and population demographics. It covers 147 countries and produces the most geographically broad view available, but it is scoped to individual users and leans toward Microsoft-adjacent tools.
The OECD ICT Access and Usage Database measures the share of firms using AI technologies, defined as tools including machine learning, natural language generation, computer vision, and speech recognition. Firms surveyed must have 10 or more employees. Coverage is limited to countries with national statistical offices capable of conducting the enterprise survey — primarily OECD members. The 2025 reference-year data (released January 2026) shows 20.2% of OECD firms using AI.
Eurostat’s ISOC_EB_AI is the EU implementation of the same enterprise survey methodology, published each December for the prior reference year. The 2025 release covers all 27 EU member states plus Norway and some candidate countries. It is the gold standard for EU enterprise-level tracking.
McKinsey’s State of AI Global Survey covers 1,993 executives across 105 countries, weighted by each nation’s GDP contribution. It asks whether organizations use AI in at least one business function. The 88% figure from November 2025 reflects this broader, self-reported framing and skews toward large organizations.
Ask three different analysts what the global AI adoption rate is, and you will get 16.3%, 20.2%, or 88% — all defensible, none comparable. The metric that matters depends entirely on the question being asked.
Which Countries Have the Highest Population-Level AI Usage?
According to Microsoft’s AI Diffusion Report H2 2025, based on anonymized telemetry across 147 countries, the top economies by working-age population AI usage in H2 2025 are:
| Rank | Country | H2 2025 Usage | H1 2025 Usage | Change |
|---|---|---|---|---|
| 1 | United Arab Emirates | 64.0% | 59.4% | +4.6 pp |
| 2 | Singapore | 60.9% | 58.6% | +2.3 pp |
| 3 | Norway | 46.4% | 45.3% | +1.1 pp |
| 4 | Ireland | 44.6% | 41.7% | +2.9 pp |
| 5 | France | 44.0% | 40.9% | +3.1 pp |
| 6 | Spain | 41.8% | 39.7% | +2.1 pp |
| 7 | New Zealand | 40.5% | 37.6% | +2.9 pp |
| 8 | Netherlands | 38.9% | 36.3% | +2.6 pp |
| 9 | United Kingdom | 38.9% | 36.4% | +2.5 pp |
| 10 | Qatar | 38.3% | 35.7% | +2.6 pp |
| 18 | South Korea | 30.7% | 25.9% | +4.8 pp |
| 24 | United States | 28.3% | 26.3% | +2.0 pp |
| — | China | 16.3% | 15.4% | +0.9 pp |
| — | India | 15.7% | 14.2% | +1.4 pp |
| — | Global average | 16.3% | 15.1% | +1.2 pp |
Source: Microsoft AI Economy Institute / AI Diffusion Report H2 2025, January 2026. Population-level tracking covers working-age adults who used a generative AI product at least once during the reported half-year period.
South Korea’s jump from 25th to 18th place is the sharpest ranking move among the top 30 economies, and it illustrates something the headline leaderboard conceals: position is not static, and the drivers of change are structural, not random. Three forces converged in Korea during H2 2025 — the government’s August 2025 AI Basic Act, GPT-5’s measurable improvement in Korean-language capability (benchmark performance on the Korean CSAT rose from the 75th to effectively the 99th percentile between GPT-4o and GPT-5), and a viral Ghibli-style image generation trend on Korean social platforms in April 2025 that pulled millions of first-time users into AI tools who then stayed. Not one of those is easily reproducible elsewhere.
The United States’ position is the most analytically interesting: 24th at 28.3% population usage, while simultaneously leading the world in AI private investment ($285.9 billion in 2025, per Stanford HAI’s 2026 AI Index), frontier model production (59 notable models in 2025, per Epoch AI), and enterprise AI adoption (88% of surveyed firms using AI in at least one function, per McKinsey). That combination — leading the world in building AI and investing in AI while ranking 24th in actually using it — is the central paradox of American AI in 2026. Leadership in infrastructure, per Microsoft’s analysis, does not automatically translate into broad population-level diffusion.
Why Do the UAE and Singapore Lead?
The UAE’s position at 64.0% is not accidental and not recent. The country appointed the world’s first Minister of State for Artificial Intelligence in October 2017 — five years before ChatGPT changed public consciousness — and launched a national AI strategy covering nine priority sectors in the same year. By the time generative AI tools arrived at scale, UAE residents had been interacting with government AI deployments in public services for half a decade. The technology felt familiar rather than novel.
According to the 2025 Edelman Trust Barometer, UAE AI trust sits at approximately 67%, compared to 32% in the United States and similar figures across Western Europe. That 35-percentage-point trust differential is one of the most underappreciated structural factors in cross-country AI adoption analysis. Populations that trust the technology enough to engage regularly produce the adoption rates that flow from repeated use, not first-time trial.
Singapore’s 60.9% reflects a different path to the same outcome: a city-state with limited geographic friction for technology rollout, consistently high digital infrastructure investment, and a skills development system that has been integrating digital competencies at scale for years. Switzerland and Singapore jointly lead the world in AI researchers and developers per capita, according to Stanford HAI’s 2026 AI Index, at 110.5 and 109.5 per 100,000 inhabitants respectively — a talent density that shapes both the supply of AI products and the population’s comfort using them.
Which Countries Lead Enterprise AI Adoption?
Enterprise adoption data from Eurostat and the OECD tells a different story from the population rankings, and the divergence is analytically significant.
EU Enterprise AI Adoption by Country, 2025
According to Eurostat’s December 2025 release (ISOC_EB_AI, 2025 reference year):
| Country | 2025 Enterprise AI Adoption | 2024 Rate | Change |
|---|---|---|---|
| Denmark | 42.0% | 27.6% | +14.5 pp |
| Finland | 37.8% | 24.4% | +13.5 pp |
| Sweden | 35.0% | 25.1% | +9.9 pp |
| Belgium | 34.5% | 24.7% | +9.8 pp |
| Netherlands | 33.2% | — | — |
| EU27 average | 20.0% | 13.5% | +6.5 pp |
| Romania | 5.2% | 3.1% | +2.1 pp |
| Poland | 8.4% | 5.9% | +2.5 pp |
| Bulgaria | 8.6% | 6.5% | +2.1 pp |
Source: Eurostat, “Use of artificial intelligence in enterprises,” December 2025. Data based on the 2025 EU ICT Enterprise Survey, covering enterprises with 10+ employees across EU member states. Dataset code: ISOC_EB_AI.
The Nordics’ dominance in enterprise adoption contrasts sharply with their middle-of-the-pack position on population usage. Denmark’s enterprises (42.0% adoption) are more active with AI than Denmark’s general population (28.7%). Axis Intelligence Research calculates this difference as an Adoption Disparity Score: Denmark scores -13.3, Finland -10.5, Sweden -1.7 — meaning enterprise adoption leads population adoption in all three. Belgium (+1.5) and the Netherlands (+5.7) show the reverse pattern, where population usage has outpaced enterprise adoption.
The structural explanation is not complicated. Nordic economies have high rates of digital infrastructure investment, enterprise technology spending, and organized training programs, all of which produce formal, measurable enterprise adoption. They also have smaller consumer markets and, in Finland’s case, a language that was historically underserved by major LLMs — a barrier that is only now receding as model multilingual capability improves.
The EU-wide 6.5 percentage point jump in one year — from 13.5% in 2024 to 20.0% in 2025 — is the largest single-year increase since EU enterprise AI measurement began. According to Eurostat, the primary reason is the spread of general-purpose generative AI tools like ChatGPT and Microsoft Copilot, which became enterprise-accessible at scale during 2024 and showed up in 2025 survey data.
OECD Firm-Level Adoption
The picture from the OECD is consistent with Eurostat but adds sectoral detail. According to the OECD ICT Access and Usage Database (January 2026), firm-level AI adoption reached 20.2% across OECD countries in 2025, up from 14.2% in 2024 and 8.7% in 2023 — more than doubling in two years.
The sector breakdown is as important as the headline:
| Sector | OECD Average Firm AI Adoption, 2025 |
|---|---|
| Information and Communication Technology | 57.3% |
| Professional, Scientific and Technical Services | 36.8% |
| Manufacturing | ~19.1% |
| Construction | ~10.8% |
| Accommodation and Food Services | Growing fastest (62.5% YoY growth) |
Source: OECD ICT Access and Usage Database, January 2026 release.
The 3× size gap between large and small firms is the most actionable finding in the enterprise data: 52.0% of large firms (250+ employees) use AI, against 17.4% of small firms. That disparity holds across virtually every country measured. It means national headline adoption rates are meaningfully determined by the size distribution of a country’s enterprise population, not just its culture or infrastructure.
The United States: AI Superpower, Middling Adopter
McKinsey’s November 2025 State of AI survey found that 88% of respondents’ organizations use AI in at least one business function — up from 78% a year earlier, surveyed across 1,993 participants in 105 nations. That figure is real but reflects a framing that deserves scrutiny. “Using AI in at least one business function” includes a single marketing team experimenting with a copywriting tool. Only one-third of McKinsey’s respondents say their organizations have moved into the scaling phase, and just 6% qualify as high performers attributing more than 5% of EBIT to AI use.
The US Census Bureau’s Business Trends and Outlook Survey, a more conservative and government-administered instrument, found AI usage hovering between 17% and 20% of all US firms between December 2025 and May 2026 — a figure consistent with the OECD’s broader 20.2% OECD-average figure.
At the population level, the 28.3% usage rate Microsoft records for the US ranks 24th globally. That ranking is not a failure of American AI development — the US leads the world in private AI investment ($285.9 billion in 2025, compared to China’s $12.4 billion, per Stanford HAI’s 2026 AI Index), in frontier model releases (59 notable models in 2025), and in AI talent (220,520 identified AI authors and inventors, per Zeki data cited in Stanford HAI 2026). It is, rather, a finding about diffusion. The US builds the tools that others in more digitally concentrated economies — UAE, Singapore, Norway — use at higher rates.
Microsoft’s analysis of its own data frames this directly: leadership in AI infrastructure and frontier model development does not by itself lead to broad population-level AI adoption.
South Korea’s Rise and What It Reveals About Adoption Drivers
South Korea’s jump from 25th to 18th in H2 2025 — a gain of 4.8 percentage points, the largest of any country in that period — is the clearest natural experiment in the dataset on what actually moves national adoption rates.
Three forces converged within a six-month window:
Policy: In September 2025, Korea enacted the AI Basic Act and reconstituted its National AI Strategy Committee as a cross-ministerial body. Simultaneously, it announced a $1.2 billion investment to integrate AI education from elementary school through the workplace, and plans to expand AI-focused science high schools in regional areas outside Seoul.
Model capability: GPT-4o, released in April 2025, achieved the 90th to 95th percentile on the Korean CSAT benchmark — the university entrance exam used as a measure of scholastic aptitude. GPT-5 pushed that to effectively the 99th percentile. For Korean speakers, this represented a step-change from a tool that was unreliable in their language to one that could handle professional, educational, and domain-specific tasks. South Korea is now the second-largest ChatGPT subscriber market globally, behind only the United States, per OpenAI’s own data cited in the Microsoft AI Diffusion Report.
Cultural moment: A Ghibli-style image generation trend using ChatGPT-4o went viral on Korean social platforms in April 2025, setting a daily ChatGPT usage record in the country. Many users who joined for the image generation feature continued using the tool for other tasks after the trend faded.
None of those three factors operates in isolation, and none is easily transplanted. But the combination reveals something important: meaningful national adoption acceleration requires simultaneous movement on institutional capacity (policy, education), technical accessibility (language support, pricing), and social transmission (viral moments that normalize the tool). Countries waiting for one of those three while ignoring the others will wait a long time.
What Is Driving the North-South AI Divide?
According to Axis Intelligence Research’s calculation based on Microsoft’s AI Diffusion Report H2 2025, the Global North’s working-age population AI adoption grew 1.8 times faster than the Global South’s during H2 2025. The North added 1.8 percentage points (from 22.9% to 24.7%) while the South added 1.0 percentage points (from 13.1% to 14.1%), widening the gap from 9.8 to 10.6 percentage points.
The gap exists for reasons that compound rather than offset each other: internet penetration, device ownership, ability to pay subscription costs in hard currency, language representation in major LLMs, and the presence or absence of local AI tools adapted to regional contexts.
DeepSeek’s rise in H2 2025 is the most significant counterweight to this trajectory. By releasing its model under an MIT open-source license and offering a free web and mobile chatbot with no subscription requirement, DeepSeek removed two of the main barriers — cost and access restrictions — that limited Western AI platform penetration in markets like Africa, where DeepSeek usage is estimated to be 2 to 4 times higher than in other regions, per Microsoft’s analysis. China, Belarus, Cuba, Russia, and Iran all show elevated DeepSeek adoption, partly reflecting geopolitical access constraints on Western platforms and partly the strategic deployment of the tool in markets where Chinese technology firms are building infrastructure partnerships.
This dynamic introduces a dimension that pure adoption rankings miss: the identity of the AI platform matters geopolitically in ways that “percentage of population using AI” does not capture. A country at 12% AI usage on DeepSeek and a country at 12% on ChatGPT report identically in the Microsoft metric but represent structurally different positions in the global AI landscape.
Axis AI Adoption Disparity Score: The Cross-Metric Gap Nobody Publishes
Most AI adoption coverage picks one metric and runs with it. The more revealing analysis compares metrics against each other, because the direction of the gap tells you something about the structure of a country’s AI relationship.
According to Axis Intelligence Research, the Axis AI Adoption Disparity Score (AAADS) is calculated as:
AAADS = ((population_adoption_rate − enterprise_adoption_rate) ÷ population_adoption_rate) × 100
A positive AAADS means individual/population usage exceeds official enterprise deployment — adoption is running ahead of formal organizational uptake. A negative AAADS means enterprise adoption leads population usage — the corporate layer is more active than the general public.
Axis Intelligence Research computed AAADS baselines for the five EU economies where both metrics are available at the country level from verified primary sources (Microsoft H2 2025 for population; Eurostat ISOC_EB_AI 2025 for enterprise):
| Country | Pop. Adoption | Enterprise Adoption | AAADS | Signal |
|---|---|---|---|---|
| Denmark | 28.7% | 42.0% | -46.3 | Enterprise leads strongly |
| Finland | 27.3% | 37.8% | -38.5 | Enterprise leads strongly |
| Sweden | 33.3% | 35.0% | -5.1 | Near parity, enterprise slightly ahead |
| Belgium | 36.0% | 34.5% | +4.2 | Near parity, population slightly ahead |
| Netherlands | 38.9% | 33.2% | +14.7 | Population leads enterprise |
Source: Axis Intelligence Research calculation. Population data: Microsoft AI Diffusion Report H2 2025. Enterprise data: Eurostat ISOC_EB_AI, December 2025.
Methodology note: Population and enterprise metrics use different denominators (working-age adults vs. firms with 10+ employees) and different definitions of “using AI.” AAADS is a directional index, not a ratio of comparable quantities. Full methodology and limitations disclosed in the Methodology section of this article.
The Nordic cluster of negative AAADS values is the most analytically interesting. Denmark and Finland have enterprise adoption rates that are 46% and 38% above their population usage rates respectively. This reflects the structure of Nordic economies — high rates of organized enterprise technology adoption driven by government programs, industry consortia, and concentrated corporate sectors — rather than a failure of consumer diffusion. The policy instruments that pushed enterprise adoption in Denmark ran ahead of the organic consumer adoption that the UAE’s trust infrastructure and Korea’s cultural moments produced.
Whether negative AAADS represents a sustainable path or a fragile one (enterprise adoption without the cultural normalization that population usage provides) is the open question. Axis Intelligence Research will track AAADS across future measurement periods to determine whether the Nordic enterprise-first model closes the population gap over time or stabilizes at the current spread.
AI Adoption by Region: Where Are the Growth Markets?
Europe: Enterprise Adoption Accelerating, Population Fragmented
The EU27 enterprise average of 20.0% in 2025 masks a 37-percentage-point spread between Denmark (42.0%) and Romania (5.2%). Eastern and Southern Europe remain substantially below the EU mean, constrained by smaller digital infrastructure investment, lower enterprise technology budgets, and, in some cases, language barriers that only recently began receding as LLM multilingual capability improved.
At the population level, France (44.0%), the Netherlands (38.9%), and the United Kingdom (38.9%) rank significantly higher than their official enterprise numbers — suggesting that in Western European consumer markets, individual AI tool adoption is running well ahead of what formal organizational surveys capture.
Asia-Pacific: Highest Usage Leaders and Fastest Growth Markets Simultaneously
The Asia-Pacific region holds both the current global leaders (UAE, Singapore) and some of the fastest growth trajectories. South Korea’s 4.8 pp gain is the largest single-half increase recorded by Microsoft in its H2 2025 dataset. Japan’s gain of 2.4 pp (to 19.1%) represents meaningful movement from a country whose AI population usage had significantly lagged its technological capability and investment. Vietnam grew 2.3 pp to 23.5%, driven partly by DeepSeek adoption and partly by a young, highly mobile-connected population.
Africa and South Asia: Still Disconnected from the Acceleration
Sub-Saharan Africa’s AI adoption figures remain in the 7-10% range across most countries measured, with DeepSeek providing the most significant recent uplift. Nigeria, Ghana, and most West African countries cluster between 8.7% and 9.3%, per Microsoft’s data. The limiting factors are structural: internet penetration, smartphone ownership, and subscription cost barriers. India at 15.7% sits just below the global average, a figure that understates the scale of India’s AI engagement in absolute terms — at its population size, 15.7% represents over 200 million individuals — while accurately capturing the diffusion challenge in reaching rural and lower-income populations.
What Does “AI Adoption” Actually Mean for Enterprise Impact?
High national adoption rates do not automatically translate into enterprise value capture. McKinsey’s November 2025 survey — 1,993 respondents across 105 countries — found that only 39% of organizations attributed any EBIT impact to AI use, and among those, most reported less than 5% of EBIT attributable to AI. Just 6% qualified as high performers, defined as organizations where AI drives 5% or more of EBIT.
The pattern McKinsey identified is instructive for interpreting national rankings: organizations that are scaling AI and capturing value are three times more likely than peers to have fundamentally redesigned workflows (not just added AI tools on top of existing processes), three times more likely to have senior leaders who demonstrate active ownership of AI initiatives, and more likely to have set growth and innovation — not just efficiency — as objectives of their AI programs. The agentic layer is where this is playing out most visibly: Axis Intelligence Research’s agentic AI statistics tracks enterprise agent deployment rates by sector and function.
This means the Denmark-vs-Romania gap in enterprise adoption (42% vs. 5.2%) describes a starting position, not an outcome. What enterprises do with AI adoption matters more than the adoption rate itself. The McKinsey high-performer cohort is distributed across countries, not concentrated in the leaders.
Barriers to AI Adoption by Country: What the Enterprise Data Reveals
Among EU enterprises that had considered using AI but decided not to, Eurostat identified the following primary barriers in 2025:
| Barrier | % of Enterprises Citing |
|---|---|
| Lack of relevant expertise | 70.9% |
| Lack of clarity on legal consequences | 52.5% |
| Concerns about data protection and privacy | 48.8% |
| Insufficient technology infrastructure | — |
| AI considered not useful for enterprise | 20.7% (least cited) |
Source: Eurostat, ISOC_EB_AI, December 2025.
Skills shortages dominate the barrier landscape across virtually every country that measures them. The OECD’s 2025 AI adoption study, drawn from a G7+Brazil survey, found similar results: skills and expertise shortfalls are the primary constraint on enterprise AI diffusion, ahead of cost, legal uncertainty, or infrastructure gaps. For the full picture on global AI adoption trends beyond country rankings, see Axis Intelligence Research’s AI statistics 2026.
This matters for interpreting national rankings. Countries that have invested heavily in AI skills — the UAE’s national AI curriculum programs, South Korea’s $1.2 billion AI education initiative, Finland’s widespread AI literacy programs — are not just addressing tomorrow’s problem. They are removing the primary barrier that prevents enterprise AI adoption from following individual AI usage at the pace the technology’s availability would otherwise allow.
AI Adoption by Sector: Where the Enterprise Gaps Are Largest
The OECD’s sectoral breakdown reveals a consistent global pattern: the information and communication sector (57.3% firm adoption) operates in a different universe from construction (approximately 10.8%) and hospitality. Within that ICT leadership, software development is the sharpest example: Axis Intelligence Research’s AI coding assistant statistics show that over 80% of new GitHub users now adopt Copilot within their first week. Professional services at 36.8%, manufacturing at approximately 19.1% — these sectoral numbers are the most actionable figures in the dataset, because national adoption averages are largely a function of each country’s industry mix.
Denmark’s 42% enterprise adoption rate is partly an artifact of Denmark having a higher share of services-sector firms than, say, Romania. Comparing raw national enterprise adoption rates without adjusting for industry composition introduces systematic bias in country-to-country comparisons.
The Eurostat sectoral data for the EU adds another layer: among enterprises using AI in 2025, the information and communication sector’s most common AI application was R&D and innovation activity (42.5%), while accommodation enterprises led with marketing and sales (58.8%). Retail used AI primarily for marketing (48.2%). The one-size AI adoption story conceals sectoral deployment patterns that are, in practice, quite different.
AI Investment and Infrastructure: The Supply Side of Adoption
Population and enterprise adoption are demand-side metrics. The supply side — where AI models are built, where the data centers sit, where investment concentrates — shapes what will be available to adopt in two to three years.
On this dimension, the US leads decisively. According to Stanford HAI’s 2026 AI Index, US private AI investment reached $285.9 billion in 2025, more than 23 times China’s $12.4 billion — part of a global AI spending picture that Axis Intelligence Research tracks separately in its AI spending statistics. The US hosts 5,427 AI data centers, more than ten times any other country (Germany is second with 529, the UK third with 523). Of 102 notable AI models released in 2025, 59 originated from US-affiliated organizations, against 35 from China.
But the US investment dominance coexists with a meaningful structural fragility: nearly every leading AI chip is fabricated by a single Taiwanese foundry, TSMC, which Stanford HAI’s 2026 Index identifies as a single point of dependency in the global AI hardware supply chain. And the pipeline of incoming AI talent to the US has fallen 89% since 2017, with an 80% decline in the single year 2024-2025 alone — the downstream consequence of which will only appear in future model production and research output, not in current adoption figures.
China leads global AI patent grants at 74.2% of the 2024 total (131,121 patents granted worldwide), and leads in AI publication volume (17.8% of all AI publications in 2024). But US patents carry significantly more forward citations — a proxy for downstream influence — despite lower volume: US-originated patents account for 51.9% of AI patent forward citations, against China’s 29.8%. Axis Intelligence Research publishes a dedicated analysis of China AI statistics covering investment, model production, and market structure in depth.
Methodology
How Axis Intelligence Research Assembled This Dataset
This report synthesizes four primary measurement frameworks, each fetched and read during production:
Microsoft AI Diffusion Report H2 2025 (January 2026): Aggregated, anonymized telemetry measuring the share of working-age adults using generative AI products at least once in H2 2025. Covers 147 economies. Adjusted for OS and device-market share, internet penetration, and country population. Retrieved from Microsoft Research, January 2026.
Eurostat ICT Enterprise Survey, 2025 reference year (December 2025): EU-statutory enterprise survey covering approximately 157,000 enterprises from a universe of 1.53 million EU firms with 10 or more employees. Surveys conducted by National Statistical Authorities in early 2025. Dataset code: ISOC_EB_AI. Retrieved from ec.europa.eu/eurostat, July 2026.
OECD ICT Access and Usage Database (January 2026): Firm-level AI adoption across OECD member countries, using comparable enterprise survey methodology. Covers the 2025 reference year. Retrieved from oecd.org, July 2026.
McKinsey Global Survey on the State of AI, November 2025: 1,993 respondents across 105 countries, weighted by GDP contribution. Online survey fielded June 25 to July 29, 2025. Not a representative sample of all businesses; skews toward larger organizations. Retrieved from mckinsey.com, July 2026.
Stanford HAI AI Index 2026 (April 2026): Ninth annual edition, synthesizing data from Epoch AI, Cloudscene, Zeki, OpenAlex, GitHub, Hugging Face, and multiple national statistical sources. Retrieved from hai.stanford.edu, July 2026.
Axis AI Adoption Disparity Score (AAADS) — Original Metric
Formula: AAADS = ((population_adoption_rate − enterprise_adoption_rate) ÷ population_adoption_rate) × 100
Inputs: Population rate from Microsoft H2 2025 AI Diffusion Report (working-age adults using gen AI at least once in H2 2025). Enterprise rate from Eurostat ISOC_EB_AI December 2025 (share of enterprises with 10+ employees using at least one AI technology in 2025).
Limitations and compatibility notes: The two inputs measure different populations (individuals vs. firms), different activities (any generative AI use vs. any AI technology deployment), and different time windows. AAADS is a directional index — it indicates whether enterprise formalization leads or trails population-level usage — not a ratio of equivalent quantities. Cross-country AAADS comparisons are valid; comparisons between AAADS and raw adoption rates require the caveat above.
Why this metric exists: No existing source compares population-level and enterprise-level AI adoption within the same country using verified primary data. Most analyses pick one metric or the other. The AAADS baseline published here is the first systematic cross-country comparison Axis Intelligence Research is aware of that does this explicitly. It will be updated in each subsequent edition as new primary data becomes available.
Calculation date: July 27, 2026. This is the baseline reading for the AAADS; the index had no prior readings.
Known Gaps and Limitations
Countries with the highest enterprise adoption (UAE, Singapore, Norway, Ireland, France) lack comparable enterprise-level data from Eurostat or the OECD because they are not EU members or are not covered in the OECD’s 2025 enterprise release at the country level. AAADS could therefore only be computed for five EU economies where both data points exist at the country level.
The Microsoft AI Diffusion metric is derived from telemetry that covers Microsoft-adjacent tools. Countries or populations with different device ecosystems or preference for non-Microsoft AI platforms — China’s domestic AI ecosystem in particular — may be undercounted.
McKinsey’s 88% figure is not directly comparable to OECD or Eurostat enterprise figures because it is self-reported, covers “any AI use in any function,” is not weighted by firm size, and samples an executive-heavy population.
About This Dataset
This dataset covers AI adoption measurements across 70+ economies and sub-groups, spanning H1 2025 through H2 2025 for population-level data and the 2025 reference year for enterprise data. Axis Intelligence Research compiled it to resolve the fragmented, methodologically incompatible state of existing AI adoption comparisons. The dataset includes population-level usage from Microsoft’s H2 2025 telemetry, enterprise adoption from Eurostat’s December 2025 ISOC_EB_AI release, OECD firm-level data from January 2026, McKinsey survey data from November 2025, and Stanford HAI contextual data from April 2026.
Axis Intelligence Research is the creator and publisher. The dataset is licensed CC BY 4.0. Cite as: “Axis Intelligence Research, AI Adoption by Country 2026, axis-intelligence.com/ai-adoption-by-country/.”
Frequently Asked Questions
Which country has the highest AI adoption rate in 2026?
The UAE leads at 64.0% of the working-age population using generative AI tools in H2 2025, per Microsoft’s AI Diffusion Report (January 2026). At the enterprise level, Denmark leads EU-tracked economies at 42.0%, per Eurostat’s December 2025 release. The answer depends on which metric you apply.
What percentage of people globally use AI?
According to Axis Intelligence Research’s analysis of Microsoft’s AI Diffusion Report H2 2025, roughly 16.3% of the world’s working-age population used a generative AI product at least once during H2 2025 — approximately one in six people. The Global North rate is 24.7%, the Global South rate is 14.1%.
Is enterprise AI adoption the same as population AI usage?
No. Enterprise adoption measures the share of firms using AI technologies in business operations. Population usage measures the share of individuals who have used an AI product. The two metrics use different denominators, different activity definitions, and often produce divergent country rankings. Countries like Denmark and Finland lead enterprise adoption but rank mid-table on population usage; countries like UAE and Ireland lead population usage but lack comparable enterprise-level data.
Why does the US rank 24th in AI usage despite leading AI investment?
According to Microsoft’s AI Diffusion Report H2 2025, the US ranks 24th at 28.3% working-age population AI usage — well below the UAE (64.0%), Singapore (60.9%), and 22 other economies. The report explicitly states that “leadership in innovation and infrastructure, while critical, does not by themselves lead to broad AI adoption.” The US leads in building and investing in AI while smaller, more digitized and AI-focused economies have driven broader population-level uptake through government skilling programs, higher baseline digital trust, and in some cases more concentrated populations where diffusion requires fewer distribution channels.
How fast is AI adoption growing globally?
Global population-level AI usage grew from 15.1% in H1 2025 to 16.3% in H2 2025 — a gain of 1.2 percentage points in six months, per Microsoft. EU enterprise adoption grew from 13.5% in 2024 to 20.0% in 2025 — a gain of 6.5 percentage points in a year, per Eurostat. OECD firm-level adoption grew from 8.7% in 2023 to 20.2% in 2025 — more than doubling in two years, per the OECD’s January 2026 data release.
Which sector has the highest enterprise AI adoption?
Across OECD countries, the information and communication technology (ICT) sector leads at 57.3% firm adoption, followed by professional and scientific services at 36.8%, per the OECD’s January 2026 data. Construction (approximately 10.8%), accommodation, and agriculture trail significantly.
Is the gap between AI leaders and laggards growing?
According to Axis Intelligence Research’s analysis of Microsoft’s H2 2025 data, the Global North-South divide widened from 9.8 to 10.6 percentage points between H1 and H2 2025. Among individual economies, the top 10 fastest-growing markets in the period were all high-income economies, meaning momentum is concentrating among already-established adopters rather than catching up the laggards.
Citation Block
APA: Axis Intelligence Research & Mitchell, S. (2026, July 27). AI adoption by country 2026: Rankings, data & the widening global divide. Axis Intelligence. https://axis-intelligence.com/ai-adoption-by-country/
MLA: Axis Intelligence Research and Sarah Mitchell. “AI Adoption by Country 2026: Rankings, Data & the Widening Global Divide.” Axis Intelligence, 27 July 2026, axis-intelligence.com/ai-adoption-by-country/.
Chicago: Axis Intelligence Research and Sarah Mitchell. “AI Adoption by Country 2026: Rankings, Data & the Widening Global Divide.” Axis Intelligence, July 27, 2026. https://axis-intelligence.com/ai-adoption-by-country/.