If your strategic AI plan treats the United States as the global leader on AI adoption, the 2025 population-level picture tells a different story. On the Microsoft AI Economy Institute and World Bank's country adoption ranking, the UAE leads at 64%. Singapore is second at 60.9%. The United States is ranked 24th, with 28.3% population-level GenAI adoption.

The framework first. The Microsoft–World Bank measure asks what percentage of a country's adult population reports using generative AI tools. It is a population-level adoption metric, not an enterprise deployment metric. It is also distinct from the McKinsey corporate adoption measure (which tracks AI use in business functions). The two measures answer different questions, and the answers diverge sharply for some countries.

The 2025 ranking surface is striking. The top four countries by population-level GenAI adoption are the UAE (64%), Singapore (60.9%), and two other gulf and Asia-Pacific economies. The traditional AI-leading countries (the United States, the United Kingdom, Germany, Japan) sit below the top tier. The United States at 28.3% places it 24th out of the 70+ countries measured. The United Kingdom and Germany rank lower still on this specific measure.

[CHART fig_4311_2026 — Population-level AI adoption by country, 2025]

The contested question this raises: what does AI leadership mean, and is the United States still leading by the measures that matter strategically?

The case for "the US is still the AI leader despite this ranking" rests on several points. The corporate adoption picture (McKinsey) shows US enterprises at 78–88% AI-in-function rates, comparable to other developed economies. US investment scale dominates the global picture ($285.88B vs $12.41B for China; see #69 The investment ramp doubled). The frontier model providers (OpenAI, Anthropic, Google, Meta, xAI) are concentrated in the US, with no comparable concentration elsewhere. By capital, by frontier model output, by enterprise AI deployment, the US remains substantially ahead.

The case for "the US is now lagging on the dimensions that matter strategically" rests on different points. Population-level adoption is a leading indicator of workforce capability: when a high share of the population has hands-on experience with GenAI, the workforce that emerges has higher baseline AI literacy. Countries with 60%+ population adoption (UAE, Singapore) are building workforce capability faster than countries at 28% (the US). The McKinsey corporate adoption metric measures whether AI is used somewhere in the organisation, not how intensively. A country with high population-level adoption may be doing more with the AI it has deployed, even if it has less of it.

The two interpretations are not mutually exclusive. The data shows the US is still leading on investment, frontier model output, and aggregate enterprise deployment. It is also showing slower population-level adoption than several smaller, more centralised economies. Both can be true simultaneously, and the strategic implications depend on which dimension the planning question is asking about.

The Anthropic Economic Index data provides another angle. Within population-level AI usage, the US shows particular concentration in computer and mathematical occupations (roughly 40% of those occupations are using AI tools). The same data shows education-sector adoption climbing from 9% early in the year to 14% twelve months later. The distribution within the US is uneven: some occupations and sectors are using AI heavily, others barely at all. The 28.3% population average masks substantial heterogeneity.

For strategic planning, three implications follow.

The first: the country-leader question depends on which dimension you are optimising for. Strategic plans that target "be in the leading geography for AI workforce capability" should be evaluating UAE, Singapore, and similar high-adoption countries, not the US. Strategic plans that target "be in the leading geography for AI capital availability or frontier model access" should still be US-centric. Strategic plans that need both, workforce capability and capital access, face an alignment problem the data has surfaced and not resolved.

The second: the geographic concentration of US AI investment (#76 AI investment concentration) interacts with the lower population-level adoption rate. The investment is concentrated in a few states (California captures 75% of US AI investment). The population-level adoption is unevenly distributed too: coastal urban populations adopt faster, interior and rural populations slower. The "US is leading" narrative obscures internal geographic variation that may matter more for planning than the national average.

The third: the ranking surface signals where to recruit and where to deploy. If the strategic question is "where can we find AI-fluent workforce most easily?", the population-level data identifies the UAE, Singapore, and similar high-adoption economies as candidates that may not have been on the recruiting map a year ago. The data does not say these countries have larger labour markets (they obviously do not), but the share of available workforce with AI fluency is higher there than in the US.

The contested question, "Is the US still the AI leader?", does not have a clean answer in the current data. By investment, by frontier model output, by aggregate enterprise deployment, yes. By population-level adoption, by workforce AI literacy, by speed of catch-up in catching-up regions, the picture is more mixed. Strategic plans that pick a single dimension to optimise for and accept the trade-offs on the others will produce better outcomes than plans that assume the US leads on every relevant dimension.

For executives setting AI direction, the planning anchor needs to be specific about which leadership dimension matters for the specific strategic question. The blanket "US leads in AI" framing is now outdated. The dimensional framing (leads on capital, lags on population-level adoption, mixed on workforce capability by region and occupation) is closer to the operating reality the current data describes.

Sources