Convergence Labs

Catch-up is not guaranteed.

For thirty-five years, inequality between countries has been falling, as poorer countries home to most of the world's people grew faster than rich ones. Transformative AI could speed that up or send it into reverse.

Convergence Labs works with governments in low- and middle-income countries to choose growth strategies that hold up however AI develops, and to build the state capacity to carry them out.

Get in touch What we do
Catching up with high-income countries Falling behind Little change High-income countries, and China from 2026

Bigger dots mean more people. Colour shows five-year growth in income per person relative to the black group: high-income countries, plus China from 2026 as a frontier AI power. Red means slower growth, not falling income. Actual to 2025, then the fast scenario, one possibility, not a forecast.

What we do

Four kinds of work, anchored in embedded teams alongside governments.

1

Embedded advisory and delivery

Small teams of advisers and operators work inside government: the centre of government, finance and economy ministries, investment agencies and the ministry responsible for AI. We start with one or two priorities the government already owns, find what is holding them back, and stay through implementation rather than stopping at a strategy.

  • AI scenarios and growth planning for the country
  • Frontier AI in daily use by senior officials
  • Delivery support on priorities the government owns
2

Research and public goods

Open tools and evidence on how transformative AI changes growth paths in low- and middle-income countries. It starts from the AGI Growth Atlas, which models 157 countries under three AI scenarios, and from the questions our country work raises.

  • Country and sector growth pathways by scenario
  • Where future demand and future jobs will come from
  • Monitoring how AI spreads in low- and middle-income countries
3

Talent and convening

Programmes that put frontier AI, and the people who understand it, close to decision makers, so that the capability stays in government after we leave.

  • AI-in-Government fellowships, from inside and outside government
  • Convenings for political leaders, funders and practitioners
4

Labs Fund

Small, fast grants of $10,000 to $100,000 to people and organisations who can do what our own teams cannot. We do not fund AI pilots, and we fund governments only through intermediaries they trust.

  • Local partners who already hold a government's trust
  • Think tanks and economists who can shift a country's debate
  • Research, writing and fellowships that build the field

Thirty-five years of catch-up

Since 1990, countries home to most of the world's people have grown faster than the United States. China went from 4% to 33% of US income per person, India from 5% to 13% and Vietnam from 6% to 20%. What comes next depends on how AI develops, and on what governments do about it.

Income per person as a share of the United States
1990 to 2025 actual, 2026 to 2040 under each AI scenario. Shown in the fast scenario, where most is at stake; it is one possibility, not a forecast. Switch to compare.

1990 to 2025: China rose from 4% to 33% of US income per person, and India from 5% to 13%. In 25 of 39 sub-Saharan African countries with data, the gap widened.

Each line is a country; thicker lines have more people, and countries above half of US income are drawn lighter. Hover to see any country. Actual: World Bank, GDP per person at purchasing power parity in constant 2021 international dollars, via Our World in Data. Scenarios: AGI Growth Atlas, 157 countries. They show what each view of AI implies, not forecasts.

The New Great Divergence

Two hundred years ago, average incomes in different countries were not far apart. Industrialisation changed that. Between 1820 and 1988, the part of global inequality that comes from differences between countries rose from a Gini coefficient of 17 to 63. Economic historians call this the Great Divergence.

Inequality between countries, 1820 to 2050
Gini coefficient of GDP per person across countries, weighted by population. Hover for values.
1820 to 1988: Milanović (2024), Table A2; dots mark the years with estimates. 1990 to 2025: World Bank data for 184 economies. 2026 to 2050: AGI Growth Atlas scenarios for 157 countries; the modest scenario tracks the no-AI line. Scenarios are not forecasts.
Inequality between countries (Gini coefficient)
  • The Great Divergence: 17 in 1820, 63 by 1988.
  • The Great Convergence: 61 in 1990, 45 by 2025.
  • By 2050 in the Atlas scenarios: 46 with no AI, 57 gradual, 79 fast.

The full chart appears on wider screens.

2037

In the fast take-off scenario, inequality between countries is back above its 1990 level by 2037, and still rising.

Transformative AI could reverse the fall. Countries with power, compute, skills and capable institutions would absorb it first, and the export routes that lifted East Asia, from assembly lines to business services, could narrow for countries still climbing. None of this is settled: how fast AI spreads, which industries a country bets on and how quickly its government learns to use frontier AI are choices. That is where we work.

Eight countries, three futures

The same scenarios, country by country. In the Atlas, how a country fares depends on its power, compute, skills, language and institutions. Governments can change several of these.

Actual No AI Gradual Fast
Income per person as a share of the United States; scales differ by panel. Sources as above.

How we work

Five commitments, in the order an engagement usually runs.

  1. 01

    We go where we are asked.

    We work with governments that request support and own the agenda.

  2. 02

    We plan for more than one future.

    Every priority has to make sense if AI stays a normal technology, and pay off more if it moves fast.

  3. 03

    We look at the jobs AI could create.

    Not only the jobs it exposes today. We ask where AI shifts demand, and which new sources of growth and work a country can capture.

  4. 04

    We work through trusted local partners.

    Often the best route into government is an organisation that already has its confidence. We back such organisations rather than compete with them.

  5. 05

    We publish what we learn.

    Tools and findings from one country are made available to every other.

Three futures we plan against

Each line shows yearly growth in income per person at the frontier, the United States, from 2026 to 2040 in one AGI Growth Atlas scenario, on the same scale in each panel.

30% a year
20262040
ModestAbout 2% a year throughout.

AI as a normal productivity tool, the view associated with Daron Acemoglu.

30% a year
20262040
GradualRising to about 15% a year by 2040.

Broad automation, with AGI around 2040, close to the central estimates of people building AI.

30% a year
20262040
FastAbout 30% a year from 2030.

A take-off, with AGI around 2030, in which today's infrastructure decides who gains first.

Convergence is the economist's word for poorer countries catching up with richer ones. It is the outcome we work for. It also describes how we work: bringing frontier AI together with growth policy and government delivery, two worlds that rarely meet.

People

Growth policy, frontier AI and company building, with long experience in governments and markets across low- and middle-income countries.

Rasmus Fonnesbæk Andersen

Rasmus Fonnesbæk Andersen, PhD

Executive Director

Led AI advisory across a 40-country portfolio at the Tony Blair Institute for Global Change, working with heads of state and ministers. Built the AGI Growth Atlas. AI, Growth and Jobs Fellow at the Africa Jobs Fund. Before that, seven years at Dalberg Advisors. PhD in political science from the University of Copenhagen.

Kartik Akileswaran

Kartik Akileswaran

Co-founder & Board Member

Co-founder of Growth Teams, which works with governments in Africa and India to find and launch new export industries that create productive jobs. A field builder and policy entrepreneur, he has spent fifteen years helping governments attract investment and remove binding constraints to growth, including with the World Bank and the Millennium Challenge Corporation, and led jobs, investment and productivity work at the Centre for Effective Governance of Indian States. Growth Teams also runs the Export Boom Atlas and the Growth Summit.

Daniel Yu

Daniel Yu

Co-founder & Board Member

Founding Partner of the Africa Jobs Fund, which aims to mobilise $100 million in philanthropic capital for companies that create high-productivity jobs in Africa, through export manufacturing and international labour mobility. Co-founder of Wasoko, the business-to-business e-commerce company serving informal retailers, which merged with MaxAB.

Convergence Labs is housed at Growth Teams.

We are talking with governments, funders and partners now.

If you work on growth, AI or government in low- and middle-income countries, we would like to hear from you.

hello@convergencelabs.org