News
Microsoft Report Says AI Adoption Is Surging, but Infrastructure and Language Gaps Persist
- By John K. Waters
- 02/18/2026
Artificial intelligence may be spreading faster than previous waves of consumer tech, but a new report from Microsoft's AI Economy Institute suggests its benefits are concentrating in a relatively small set of countries, with infrastructure and language emerging as major dividing lines.
The report ("AI Diffusion Report: Where AI is most used, developed, and built") estimates that more than 1.2 billion people have used AI tools in less than three years, a pace it compares with earlier general-purpose technologies. It also argues that rapid headline growth masks basic constraints: "With more than 1.2 billion users in under 36 months, AI has become the fastest-adopted technology in human history," the report argued.
Measured by the share of working-age adults using AI tools, the United Arab Emirates ranked first at 59.4%, followed by Singapore at 58.6%, Norway at 45.3%, and Ireland at 41.7%. The United States was listed at 26.3%, while China was listed at 15.4%.
The institute says its diffusion estimate relies in part on Microsoft's view into software usage. "By analyzing aggregated and anonymized telemetry from over one billion Windows devices, we can estimate the prevalence of AI-related activity across regions," the report claimed, adding that it adjusts for the fact that the dataset excludes non-Windows devices.
A related Microsoft Research technical report ("Measuring AI Diffusion: A Population-Normalized Metric for Tracking Global AI Usage") describes a population-normalized usage metric built from telemetry and adjusted for device access and mobile scaling across 147 economies.
Even with connectivity rising globally, the report's findings fit a broader pattern of digital gaps. The International Telecommunication Union estimates 5.5 billion people were online in 2024, but says about one-third of the world remains offline, with the hardest-to-connect populations concentrated in lower-income and rural regions.
Power access is another constraint the report highlights as foundational for data centers and daily AI use. The World Bank's electricity access indicator puts global access above 90% in recent years, but with much lower coverage in low-income economies and parts of Sub-Saharan Africa. The Microsoft report frames the consequence in geographic terms, saying, "AI adoption in the Global North is approximately 23%, compared with only 13% in the Global South."
On the supply side, the report argues that the compute required to build and run advanced AI remains concentrated. It says, "Datacenter capacity remains heavily concentrated, with the United States and China accounting for roughly 86% of global compute," and provides a markdown citing International Energy Agency estimates, listing 53.7 gigawatts for the United States and 31.9 gigawatts for China.
The IEA has also warned that AI is set to sharply increase electricity demand from data centers over the coming years, intensifying pressure on grids in the largest data center regions.
The report also tries to separate countries that build frontier models. It says only seven countries host "frontier-level" AI models and that the performance gap is narrowing. In a table comparing each country's best model against the frontier, it lists the United States at 0 months to frontier, China at 5.3 months, South Korea at 5.9 months, France at 7.0 months, the United Kingdom at 7.7 months, Canada at 7.8 months, and Israel at 11.6 months.
Among the report's more pointed claims is that language can be a standalone barrier, even after accounting for income and connectivity. "Countries where low-resource languages are predominant exhibit significantly lower AI adoption, even after controlling for GDP and internet access," the report asserted.
That argument offers evidence that widely used web corpora used in AI development are heavily skewed toward a small set of languages, with Common Crawl's language statistics showing a large concentration in top languages such as English.
The institute's bottom line is that diffusion is not just a question. Ultimately, the value of artificial intelligence will be judged not by the number of models produced, but by the extent to which they benefit society," the report concluded.
About the Author
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].