Key findings

  • If the pace measured from H1 2025 to Q2 2026 holds, 23.8% of people aged 15 to 64 worldwide would use generative AI by late 2027, against 18.8% now. If the pace halves, 21.3%. Either way, most of the world would still not be using it.
  • Economies with their own published value above 50% would go from 2 to 11 on trend (6 if the pace halves).
  • South Korea is the fastest mover, at 13.3 points a year. On trend it would reach 60.5% and rise from 12th to 5th. Japan would climb from 45th to 34th.
  • The leaders are pulling away: an economy's current level and its pace correlate at 0.81. The Global North-South gap would grow from 12.6 points to about 17.4 on trend.
  • Countries with more AI-exposed workforces are growing faster (r = 0.59), but that is income: holding GDP per capita fixed, the link disappears (r = -0.05).

How we extended the trend

Microsoft's AI Economy Institute publishes the share of people aged 15 to 64 in each economy who used a generative AI product, for four periods: H1 2025, H2 2025, Q1 2026 and Q2 2026. We explain that measure in Microsoft's AI Diffusion report, explained. For each economy we fitted a straight line through those four points to get its pace in percentage points per year, then extended it 18 months from mid-2026 to the last quarter of 2027.

We show two scenarios. On trend assumes the recent pace continues. Slower assumes it halves from now on. Neither is a forecast: there is no model of what drives adoption here, only an extension of a short, measured series. We only rank the 102 economies that have their own published value; the other 45 share an identical value with two or more others, which looks like a regional estimate.

The world in 2027

AI use share by region, observed and extended to late 2027 Line chart. Observed values to Q2 2026: Global North 28.8 percent, World 18.8, Global South 16.2. Dashed lines extend the recent pace to Q4 2027: Global North 38.7, World 23.8, Global South 21.3. Dotted lines show the pace halved: 33.8, 21.3 and 18.7. 10% 20% 30% 40% trend extension H1 2025 Q2 2026 Q4 2027 Global North 33.8-38.7% World 21.3-23.8% Global South 18.7-21.3%
Share of people aged 15 to 64 who used a generative AI product. Solid lines are Microsoft's measured values (World from H1 2025; North and South from H2 2025). Dashed lines extend each series' recent pace to Q4 2027; dotted lines halve it. Our calculation from Microsoft's regional tables.

World use has grown by about 3.3 points a year. Carried forward, that takes it from 18.8% in Q2 2026 to 23.8% by late 2027. The Global North is growing about twice as fast as the Global South (6.6 against 3.4 points a year), so the gap between them would widen from 12.6 points to 17.4, or 15.1 if growth halves. The regional pace rests on only three periods, so treat it as a direction, not a number to plan around.

Who would pass 50%

AI use share in Q2 2026 and late-2027 range by economy Horizontal bars. Solid bar is the Q2 2026 value; the lighter extension runs to the late-2027 value if the pace halves, and the outline to the value if the recent pace holds. United Arab Emirates 73.3 now, 82.7 to 92.2 by 2027; Singapore 64.3 now, 68.2 to 72.2 by 2027; France 49.6 now, 55.5 to 61.5 by 2027; Ireland 49.9 now, 55.5 to 61.2 by 2027; South Korea 40.6 now, 50.6 to 60.5 by 2027; Norway 49.4 now, 52.3 to 55.1 by 2027; Qatar 43.5 now, 48.8 to 54.1 by 2027; United Kingdom 43.7 now, 48.7 to 53.7 by 2027; Netherlands 43.5 now, 48.4 to 53.4 by 2027; New Zealand 44.1 now, 48.5 to 52.9 by 2027; Spain 45.1 now, 48.8 to 52.5 by 2027; Belgium 40.3 now, 45.0 to 49.6 by 2027; Australia 40.7 now, 44.9 to 49.2 by 2027; Switzerland 39.1 now, 43.7 to 48.3 by 2027; Sweden 37.3 now, 41.5 to 45.7 by 2027; United States 33.0 now, 37.5 to 42.1 by 2027; Germany 32.3 now, 36.3 to 40.2 by 2027; Japan 24.7 now, 30.1 to 35.4 by 2027; India 18.5 now, 21.4 to 24.4 by 2027; China 17.5 now, 18.7 to 20.0 by 2027. 0% 20% 40% 60% 80% 100% United Arab Emirates 73.3 → 82.7-92.2 Singapore 64.3 → 68.2-72.2 France 49.6 → 55.5-61.5 Ireland 49.9 → 55.5-61.2 South Korea 40.6 → 50.6-60.5 Norway 49.4 → 52.3-55.1 Qatar 43.5 → 48.8-54.1 United Kingdom 43.7 → 48.7-53.7 Netherlands 43.5 → 48.4-53.4 New Zealand 44.1 → 48.5-52.9 Spain 45.1 → 48.8-52.5 Belgium 40.3 → 45.0-49.6 Australia 40.7 → 44.9-49.2 Switzerland 39.1 → 43.7-48.3 Sweden 37.3 → 41.5-45.7 United States (#20) 33.0 → 37.5-42.1 Germany (#25) 32.3 → 36.3-40.2 Japan (#34) 24.7 → 30.1-35.4 India (#63) 18.5 → 21.4-24.4 China (#71)* 17.5 → 18.7-20.0
Solid bar: measured AI use share in Q2 2026. Light bar: late 2027 if the pace halves. Outline: late 2027 if the recent pace holds. Top 15 economies on trend, then selected others with their 2027 rank. Dashed line marks 50%. *China's figure is likely understated: Microsoft says the measure omits several widely used domestic tools. Source: Microsoft AI Diffusion dataset (MIT licence), our trend extension.

Today only United Arab Emirates and Singapore are above 50% among economies with their own value. On trend, 11 would be by late 2027: United Arab Emirates, Singapore, France, Ireland, South Korea, Norway, Qatar, United Kingdom, Netherlands, New Zealand and Spain. If the pace halves, the list shrinks to 6.

The UAE's line would take it to 92.2%, which is where a straight line is least believable. Adoption usually slows as it approaches saturation, so for economies already above about 60%, the slower scenario (82.7% for the UAE) is the more plausible of the two.

Who is climbing and who is slipping

South Korea has the fastest pace of any economy, 13.3 points a year, after rising from 25.9% in the first half of 2025 to 40.6% in Q2 2026. On trend it would reach 60.5% and rank 5th, up from 12th. Japan starts lower but moves almost as fast as France and Ireland (7.2 points a year), which would lift it from 45th to 34th. We looked at last year's moves in who climbed and fell in AI use rankings.

Spain shows the other side. It is 6th today at 45.1%, but its pace (5.0 points a year) is slower than most of its peers, so it would slip to 11th. A slower climb is still a climb: Spain would reach 52.5% on trend.

The median economy gains 4.0 points a year. At the other end, Ukraine, Syria and Jamaica each recorded at least one period where measured use fell, and several low-use economies are growing by about one point a year.

The largest economies

EconomyQ2 2026Pace (pts/yr)2027 slower2027 on trendRank now → 2027
United States33.0%6.037.5%42.1%21 → 20 (+1)
China17.5%1.618.7%20.0%67 → 71 (-4)
India18.5%3.921.4%24.4%64 → 63 (+1)
Japan24.7%7.230.1%35.4%45 → 34 (+11)
Germany32.3%5.336.3%40.2%=23 → 25 (-2)
United Kingdom43.7%6.748.7%53.7%8 → 8 (0)
France49.6%7.955.5%61.5%4 → 3 (+1)
Brazil20.2%4.223.3%26.4%58 → =57 (+1)
Indonesia14.8%2.816.9%19.0%76 → 75 (+1)
Mexico21.4%4.324.6%27.8%55 → 53 (+2)
South Korea40.6%13.350.6%60.5%12 → 5 (+7)
Italy31.4%5.135.2%39.0%=25 → 27 (-2)
Canada38.5%4.541.9%45.3%16 → =16 (0)
Australia40.7%5.644.9%49.2%11 → 13 (-2)
Spain45.1%5.048.8%52.5%6 → 11 (-5)
Saudi Arabia31.4%6.936.6%41.8%=25 → 21 (+4)
Turkiye18.9%5.022.7%26.4%62 → =57 (+5)
Poland32.3%5.336.3%40.3%=23 → =23 (0)
Russia9.9%2.211.5%13.2%=95 → 94 (+1)
Egypt15.5%2.717.5%19.6%=72 → 72 (0)

Ranks among the 102 economies with their own published value; "=" marks a tie. Nigeria is not shown because their value is shared with other economies or missing. Full table: ai_adoption_2027.json.

The United States would reach 42.1% on trend and stay near 20th (21st now, 20th in 2027). India would rise from 18.5% to 24.4%. China has the slowest measured pace of the large economies (1.6 points a year), but Microsoft says its measure leaves out several popular domestic tools, so China's real level and pace are probably higher than shown.

Does job exposure predict faster adoption?

We compared each economy's pace with its workforce's AI exposure, our model-estimated score computed on the nine ISCO-08 major occupation groups so that every country is measured at the same level of detail (89 economies have both). Economies whose jobs are more exposed to AI are adopting it faster: the correlation is 0.59. But richer economies have both more exposed jobs and faster adoption (exposure and income correlate at 0.89). Holding GDP per capita fixed, the link between exposure and pace falls to -0.05, which is no link at all.

So, on this data, a country's job mix does not tell you how fast it will adopt AI once you know how rich it is. That matches what we found for the level of use in AI exposure vs actual AI use. It also means that by 2027 the workers most likely to be using AI are in high-income economies, whatever their occupation.

What could break the trend

What this means for workers

Use is not displacement. A rising share of people trying AI tools says nothing direct about job losses, and none of these numbers measure them. What the trend does suggest is that by late 2027 about half or more of working-age people would be using AI in the leading economies, around two in five in the United States and Germany, and fewer than one in four across the Global South. For workers in the most exposed occupations, such as clerical support, see how your country's job mix compares in the US, the UK or the US against the world.

We will check this

These are extensions we can test. When Microsoft publishes data for the end of 2027, we will compare it with both scenarios here, economy by economy, and publish the result, including where we were wrong. Expect Microsoft's planned method change to make that comparison harder, and we will say so when it does.

See AI exposure for any country

Compare occupation-level AI exposure for over 200 countries, built from official ILO and national statistics.

Open the explorer →

Was this analysis useful?

Methodology and sources. Country values are the "AI Diffusion" columns (H1 2025, H2 2025, Q1 2026, Q2 2026) in the Microsoft AI Economy Institute's public dataset, Q2 2026 update (MIT licence): the estimated share of people aged 15 to 64 who used a generative AI product in the period. Cite: Misra, A., Wang, J., McCullers, S., White, K. and Lavista Ferres, J. (2025), Measuring AI Diffusion: A Population-Normalized Metric for Tracking Global AI Usage, doi.org/10.48550/arXiv.2511.02781. Dataset: github.com/microsoft/ai-diffusion-report. Regional values (World, Global North, Global South) are from Microsoft's report tables; the World value for H1 2025 is from the H2 2025 report text. Our calculation: each period is placed at its midpoint (H1 2025 = 2025.25, H2 2025 = 2025.75, Q1 2026 = 2026.125, Q2 2026 = 2026.375); the pace is the least-squares slope in points per year; the late-2027 value is the Q2 2026 value plus the pace times 1.5 years (on trend) or half the pace times 1.5 years (slower), capped at 95%. Rankings use only economies whose Q2 2026 value is not identical to two or more others. Exposure comparison: employment rolled up to the nine ISCO-08 major groups (armed forces excluded) and weighted by our major-group AI exposure scores; countries without all nine groups are left out; income is World Bank GDP per capita (current US$), log scale. All figures from site/data/ai_adoption_2027.json, built by scripts/build_ai_adoption_2027.py on 2026-11-11.

AI exposure, robotics, offshoring and work-from-home scores are model-estimated: an AI language model (Claude Opus 4.7) scored each ISCO-08 occupation group on 28 May 2026, guided by published research. They are informed judgements, not measurements, and not forecasts of job losses. Read the full methodology and how the scores were checked.

Frequently asked questions

How many people will use AI by 2027?
If the pace measured from early 2025 to mid-2026 holds, about 23.8% of people aged 15 to 64 worldwide would use generative AI by late 2027, against 18.8% in Q2 2026. If the pace halves, about 21.3%.
Which countries will use AI the most in 2027?
On the current trend the UAE (92.2%), Singapore (72.2%), France (61.5%), Ireland (61.2%) and South Korea (60.5%) would lead. These are trend extensions of Microsoft data, not forecasts.
Is the global AI gap getting bigger?
Yes, on current trends. The Global North-South gap in AI use grew to 12.6 points in Q2 2026. Extending each region's pace gives about 17.4 points by late 2027, or 15.1 if growth halves.
How was the 2027 AI adoption projection made?
We fitted a straight-line pace to four periods of Microsoft AI Diffusion data per economy (H1 2025 to Q2 2026) and extended it 18 months. A second scenario halves the pace. It is a trend extension, not a forecast model.

Sources

  • Microsoft AI Economy Institute - AI Diffusion dataset, Q2 2026 update, MIT licence (github.com/microsoft/ai-diffusion-report)
  • Microsoft AI Economy Institute (2026) - Global AI Diffusion reports for H2 2025, Q1 2026 and Q2 2026
  • Misra, Wang, McCullers, White and Lavista Ferres (2025) - Measuring AI Diffusion: A Population-Normalized Metric for Tracking Global AI Usage (arXiv:2511.02781)
  • World Bank - World Development Indicators, GDP per capita (current US$)
  • WorldJobsData (2026) - Occupation-level AI exposure scores, methodology; trend extension in ai_adoption_2027.json

Spotted an error or know a better source? Report it and we will check it against the source data.