AI's riches are accruing to the rich: North America, Europe and parts of Asia–Pacific lead in connectivity, education and AI deployment while large swaths of Africa and South Asia lack the infrastructure and skills to share the benefits; without concerted investment in inclusion and governance, the global AI divide—and its economic and social costs—will deepen.

Main Finding

Global adoption of AI is highly uneven: high‑income economies (North America, Europe, parts of Asia‑Pacific) are far ahead in access and use, while many low‑income countries (notably parts of Africa and South Asia) lag across infrastructure, skills and institutional capacity. Without targeted policy intervention, the AI adoption gap between the Global North and Global South is likely to widen, amplifying economic and social inequalities.

Key Points

• Magnitude of the gap
  • Worldwide: ~16.1% of working‑age people used AI in H2 2025 (Microsoft).
  • Large country variation: UAE ~64% (top), Cambodia ~5.1% (bottom); Ukraine ~9%.
  • Regional averages: Global North 24.7% vs Global South 14.1%; gap increased from 9.8 to 10.6 percentage points.
  • Growth concentrated in high‑income countries: top growth rates are almost exclusively in high‑income economies (e.g., South Korea, UAE >4%).
• Multi‑layered prerequisites for AI diffusion (pyramid)
  • 2024 world population ~8.1 bn; ~7.4 bn with electricity; ~5.5 bn with internet access; ~4.2 bn with basic digital skills; ~1.2 bn active AI users.
  • Bottlenecks: energy and telecom infrastructure, computing hardware and data centers, cloud access, data quality and governance, human capital, and regulatory capacity.
• Drivers of uneven adoption
  • Economic resources (GDP per capita, investment), digital infrastructure, education/digital literacy, state programs and industrial strategy, and private sector/tech firm activity.
• Risks and externalities
  • Widening digital divide and economic isolation for low‑adoption regions.
  • Increased global inequality and within‑country displacement risks concentrated among low‑skill workers.
  • AI bias, governance shortfalls, environmental pressures from data centers and compute demand.
• Potential for positive impact
  • AI can reduce inequality if deployed for public goods (health, education, precision agriculture), but requires deliberate policy, capacity building and inclusive design.

Data & Methods

• Data sources (secondary):
  • Microsoft AI adoption and AI Diffusion Report (2025)
  • VisualCapitalist mapping of AI adoption by country
  • World Bank Digital Progress & Trends (2025)
  • UNESCO Recommendation on the Ethics of AI
  • UNCTAD Technology & Innovation Report (2025)
  • ITU Annual AI Governance Report (2025)
  • WEF analyses, CGDev commentary and other international reports cited in the paper
• Methods used in the study:
  • Comprehensive literature and policy review of international organization reports.
  • Descriptive, cross‑country comparisons of AI adoption rates and growth (country rankings and regional aggregation: Global North vs Global South).
  • Simple visualizations (author’s figures/pyramids) to illustrate layered prerequisites and percentage comparisons.
• Limitations noted by the author:
  • Heavy reliance on secondary, rapidly changing sources and on measures of "AI use" that vary in definition across datasets.
  • Lack of extensive country‑level time series or causal econometric analysis; many indicators are snapshots rather than longitudinal measures.
  • Data gaps for many low‑income countries and potential measurement error in self‑reported/usage metrics.

Implications for AI Economics

• Distributional consequences
  • AI is likely to be a growth and competitiveness amplifier for countries already well‑positioned; this can lead to persistent cross‑country income divergence unless countered by policy.
  • Within countries, AI adoption without complementary policies risks exacerbating wage inequality and structural unemployment among lower‑skilled workers.
• Policy priorities to influence economic outcomes
  • Infrastructure: invest in reliable electricity, broadband, and affordable cloud/compute access (including regional data centers) to lower entry barriers.
  • Human capital: scale digital literacy, STEM and AI‑specific training, and reskilling programs targeted at vulnerable worker groups.
  • Institutional capacity and governance: support data governance, AI safety/ethics frameworks, and regulatory capacity in developing countries to manage risks and attract investment.
  • International cooperation and technology transfer: multilateral programs, finance and public–private partnerships to accelerate diffusion in lower‑income countries.
  • Inclusive design & public good deployment: promote AI applications in health, education and agriculture to realize pro‑poor gains.
  • Redistribution & social policy: consider progressive taxation, redistribution and social protection to manage transitional inequality.
• Research implications
  • Need for standardized metrics of AI adoption and use (to improve comparability).
  • Demand for causal empirical work on AI’s impact on growth, employment and distribution at country and regional levels.
  • Evaluation of policy experiments (infrastructure investment, training programs, governance interventions) to identify scalable, cost‑effective approaches for inclusive AI diffusion.

Overall, the paper argues that AI’s economic benefits will not be automatic or evenly distributed. Economic policy choices—domestic and international—will critically shape whether AI deepens existing inequalities or becomes a tool for broader, inclusive development.