About
About the AGI Growth Atlas
The AGI Growth Atlas is an independent, non-commercial research project. It asks one question: if artificial general intelligence arrives, how could economic growth change across 157 countries to 2060? It answers with three selectable scenarios rather than a forecast, and it attaches no probability to any of them.
Author
The atlas is created and maintained by Rasmus Fonnesbæk Andersen, a political scientist and development economist who has advised governments and organizations on technology, growth and public policy in more than 40 countries. It grew out of his work on the implications of continued AI capability growth for jobs and economic development, with particular attention to Africa and other low- and middle-income countries. It is published in a personal capacity and represents no institution.
How to cite
Andersen, R. F. (2026). AGI Growth Atlas: how AGI could reshape global growth to 2060. Version 3 (September 2026), data as of 28 September 2026. agigrowth.org
Every view has its own address, so a country page or a specific scenario view can be cited directly, for example agigrowth.org/countries/kenya?lens=fast. The country briefs are also available as PDFs from each country page.
Built with Claude
The AGI Growth Atlas has been built in collaboration with Claude, Anthropic’s AI model, since the first version in November 2025. Claude has written the site’s code and data pipelines, helped find and check sources, and drafted the 157 country briefs from the Atlas’s data. The overarching vision, questions, framework, scenarios, and analysis are Rasmus Fonnesbæk Andersen’s, as is the responsibility for any errors.
Thanks
Many people in government, development organizations, research and the AI community have commented on the atlas since its first version. Their feedback has made it clearer and more accurate, and it is gratefully acknowledged. Any remaining errors are the author’s own.
What changed in version 3 (September 2026)
- New default scenarios. Gradual now reaches 15% a year for the best-positioned country from 2040 (previously 20% from 2045); Fast reaches 30% from 2030 (previously 40%). Both ceilings now match published forecasts, and AI’s contribution to growth ramps up from 2025 rather than 2027. All of these remain adjustable in the .
- New ways in. A ten-minute visual essay (), a page for each region, a page tracking the signals that would tell the scenarios apart, and an page where you can pin up to five countries on the map and compare them side by side.
- Inequality between countries. The framework now continues Milanović’s two-century series of between-country inequality to 2060 under each scenario.
- Fresh data. Levels are 2025 actuals and each country’s no-AI trend is now its 2016–2025 average. The factor scores use the latest World Bank and governance data, the 2025 English Proficiency Index and Microsoft’s AI diffusion figures for mid-2026.
- Country archetypes in two tiers. The first tier names positions in the AI supply chain, including a new group of Regional leaders such as India, Nigeria and Brazil. The second sorts the hundred-plus countries that most writing on AI middle powers leaves out by how AI reaches them: offshored work, tourism and diaspora earnings, smallholder farms, or adoption at home. Military governments now count as High-risk.
- Levels, not only rates. GDP in 2045 against the no-AI path and average growth 2025–2045 sit beside every growth rate, and every view has its own shareable address.
The country briefs are being redrafted for the new defaults and data; until then each brief is marked as prepared under the earlier ones.
Versions and updates
The first version was published in November 2025, the second in July 2026 and the third in September 2026. Data sources are checked for updates on the first of every month; the last check was 1 October 2026 and the last data refresh 28 September 2026. Sources, scores and the model are documented in the , and the signals that would tell the scenarios apart are tracked on .