Generative AI is spreading faster across the global economy, but not evenly. In Microsoft’s latest Global AI Diffusion Report, published May 7, the company said AI usage rose to 17.8% of the world’s working-age population in the first quarter of 2026, up from 16.3% in the prior quarter. For all the noise around model launches and chip spending, that figure matters because it tries to measure actual use rather than venture funding or benchmark scores (Microsoft, May 7).
Adoption is broadening, but leadership is concentrated
Microsoft said 26 economies now have AI usage above 30% of their working-age population. The UAE remained first on its national leaderboard at 70.1%, while the United States moved up from 24th to 21st place with a 31.3% usage rate. The company also pointed to faster adoption in South Korea, Thailand and Japan, arguing that better support for Asian languages is now translating into more real-world use (Microsoft, May 7).
That matters because the measurement is unusually specific. In a November 2025 paper on arXiv, the report’s authors said their “AI User Share” metric tracks active use across 147 economies and adjusts for device access, mobile scaling, internet penetration and population. The same paper also makes the limitation explicit: because the dataset is built from anonymized Microsoft telemetry, it is not a neutral census of all AI activity, but a directional measure of diffusion that can still be useful for policy and benchmarking (Misra et al., arXiv, Nov. 2025).
The gap between North and South is getting harder to ignore
Microsoft’s most important number may be the regional split. It said AI usage in the Global North reached 27.5% in the quarter, versus 15.4% in the Global South. That is a 12.1-percentage-point gap at a moment when governments are racing to frame AI as general-purpose infrastructure rather than just a software feature (Microsoft, May 7).
The contrast is even sharper when set against the money pouring into the sector. Stanford’s AI Index 2026, summarized by IEEE Spectrum, said global AI investment hit more than $581 billion in 2025, up from $253 billion in 2024, with more than $344 billion of that total flowing into the United States alone. In other words, capital is scaling far faster than mass adoption, and the places attracting most of the money are not automatically the only places where demand exists (IEEE Spectrum, May 2026).
Coding is becoming the clearest economic test case
The report also suggests that software is where AI’s labor impact is becoming easiest to observe. Microsoft said global git pushes rose 78% year over year, while U.S. software developer employment reached about 2.2 million in 2025, up 8.5% from a year earlier. Early 2026 data cited in the report showed March employment still running about 4% above March 2025 levels (Microsoft, May 7).
That does not settle the jobs debate. But it does weaken the simplest assumption that better coding models immediately shrink demand for programmers. For now, the more plausible story is that cheaper software production is widening the set of things companies are willing to build. The next question is whether the same dynamic can narrow the global adoption gap before AI access hardens into a new kind of digital inequality.
Sources: Microsoft Global AI Diffusion Report (May 7, 2026); Misra et al., “Measuring AI Diffusion” (arXiv, Nov. 2025); Stanford AI Index 2026 as summarized by IEEE Spectrum.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.