Only 18.8% of Working-Age People Use AI. The Other 81.2% Are Watching From a Different World.

Microsoft’s AI Economy Institute puts generative AI adoption among people aged 15 to 64 at 18.8% in Q2 2026. Q1 was about 17.8%. That is a 1-point rise. With about 5 billion working-age people, roughly 940 million used generative AI in the past month. The study measures usage, not opinion. It uses Microsoft’s aggregated anonymous telemetry and adjusts for operating system, device share, internet penetration, and population size.

1. UAE 70.1%, Singapore 64.3%

The UAE sits at about 70.1%. Singapore is at about 64.3%. In the UAE, about 7 in 10 working-age people use AI. In Singapore, about 6.4 in 10. Ireland is at 49.9%. France is at 49.6%. Norway is close to 50%.

The populations differ. The UAE has about 10 million people. Singapore has about 6 million. France has about 68 million. All four have strong network coverage, high smart device use, digital government services, and multilingual talent. Microsoft does not claim cause and effect. It records adoption. The UAE and Singapore kept their lead from Q1 to Q2. Their rankings did not change.

2. South Korea: 37.1% to 40.6%

South Korea went from 37.1% in Q1 to 40.6% in Q2. That is a 3.5-point jump. Its global rank rose from 16th to 12th. Over 12 months, its adoption rate grew by about 15 points. Microsoft says South Korea had the largest increase among major economies for three quarters in a row.

Saudi Arabia rose from 30th to 25th. Japan went from 22.5% to 24.7%. Japan had the largest relative increase for itself. The three countries grew in different ways. South Korea used its consumer electronics base, local-language models, and fast corporate uptake. Saudi Arabia used national strategy and investment. Japan caught up with localized tools. Japan’s rate is still lower than South Korea’s. Its relative increase was the largest, which shows how fast a low base can move.

3. Global North 28.8%, Global South 16.2%

The Global North averages 28.8%. The Global South averages 16.2%. The gap is 12.6 points. Last quarter it was 12.1 points. The North grew 1.3 points. The South grew 0.8 points. The gap widened by 0.5 points.

The report lists the South’s obstacles: electricity supply, internet connectivity, device penetration, and digital skills. These are hard limits. No stable electricity, no local deployment. No cheap internet, high API costs. No devices, fewer use cases. No digital skills, and people open the tool and stop. The South is 13.8 points away from 30%. The North is already near 30%. The South is still in the 16% range.

4. United States 33.0%, 21st

The U.S. is at 33.0%. It ranks 21st. The UAE is 37.1 points higher. Singapore is 31.3 points higher. The U.S. has the densest cluster of AI model companies and capital. Its working-age adoption rate still ranks 21st.

The report does not explain every reason. The ranking shows that model supply and daily use are separate. Income, education, industry, firm size, privacy trust, and tool fragmentation all shape use. The U.S. market is large, diverse, and fragmented. Regulation is contested. 33.0% means about one-third have used generative AI. About two-thirds have not. That is lower than the UAE, Singapore, Ireland, France, Norway, and South Korea.

5. 26 Economies Above 30%

Twenty-six economies now have adoption above 30%. That fact shows the uneven spread better than 18.8% does. Above 30%, AI enters routine work in some markets. Most economies are still below 30%. The Global South averages 16.2%. It is 13.8 points from 30%.

The 26 economies cluster in North America, Europe, the Gulf, and East Asia. Most of Africa, South Asia, and Latin America are outside that group. The report does not list all 26. The top names include the UAE, Singapore, Ireland, France, Norway, and South Korea. The U.S. is at 33.0%, just above 30%, but ranks 21st.

6. Local Languages, Open Weights, Infrastructure

The report says local-language models in Asia are getting better. Open-weight models lower costs and make localization easier. DeepSeek has appeared in developing countries. Microsoft plans to invest $50 billion in AI in the Global South over 10 years.

These are supply-side moves. Demand still depends on electricity, internet, devices, and skills. If the money goes there, the adoption curve changes. If it goes only to models and cloud services, leading markets speed up and trailing markets wait longer. Open-weight models matter because they allow local deployment and fine-tuning without paying high API fees for every call. For the Global South, that is one way to lower the barrier.

Conclusion

Three numbers at the end of Q2: North 28.8%, South 16.2%, gap 12.6 points.

If the North keeps growing 1.3 points per quarter and the South keeps growing 0.8 points, next quarter would be North 30.1%, South 17.0%, gap 13.1 points. The report did not make that forecast. It only gave the Q2 numbers.