The low, low cost of ending extreme poverty

Ending extreme poverty could be achieved for about $318 billion per year, or roughly 0.3 percent of global GDP, through targeted direct cash transfers, according to a new analysis by anti-poverty researchers that uses AI to map the picture on the ground. The study argues the price tag is small enough to be politically and practically feasible, challenging the notion that the world’s worst poverty is too big to fix.

The researchers say their approach relies on data national governments already collect to estimate how much each family would need to cover basics — food, shelter, medicine — and then scales those needs through cash transfers. They anchor their calculations to the World Bank's threshold, which was raised this year from 2.15 to 3 dollars per day, and estimate that providing cash transfers at or near that level could lift hundreds of millions of people out of life-threatening poverty — roughly 8 percent of the world’s population. While the plan would be logistically challenging, the authors argue it would be implementable using existing administrative systems and the power of AI-informed analysis.

What makes this estimate different from earlier ones is that it isn’t a poverty gap thought experiment that promises to fill every poor household to the exact poverty line. In that model, a family earning 1 per day would get exactly 1 more, and neighbors would receive proportionally small amounts, a calculation that is cheap on paper but depends on highly granular, unavailable data. Blumenstock said that approach is not actually feasible and not shovel-ready. By contrast, the new study combines available government data with AI tools to produce a transfer plan that is more implementable in practice, even if it means paying a bit more per person.

The researchers acknowledge the money is only part of the challenge. What’s really striking is that it’s not that much money, Blumenstock said. Extreme poverty exists not because it’s prohibitively expensive to address, but because it is institutionally and politically difficult to get people to prioritize it. The note comes as rich countries have shown signs of deprioritizing affordable anti-poverty programs abroad, a shift that has already had consequences for vulnerable populations. Yet the study also notes real progress: the number of people living on less than 3 per day has fallen from more than 40 percent of the world in 1990 to around 10 percent today, even as the global population grew by more than 2.5 billion. Much of the remaining poverty is concentrated in sub-Saharan Africa, where growth has slowed and population growth remains high.

Ending extreme poverty, the authors argue, does not require miracle funding, but it does demand sustained political commitment and new approaches. The remaining pockets of extreme poverty are unlikely to vanish in the near term without deliberate policy choices, particularly in sub-Saharan Africa where economic growth has lagged. If nothing changes, population growth could push the number of extremely poor people higher in the coming years even as global wealth grows.

The study frames finishing the job as historically plausible: 35 years ago, preventing extreme poverty seemed almost as impossible as eradicating it now appears achievable at scale. The price tag of about 318 billion a year is modest compared with other global expenditures, and, the authors say, it could be funded by a mix of government budgets, philanthropy, and donor funds. The key takeaway is not merely a number but the call to treat poverty as solvable with practical tools, data, and political will.

If adopted, the plan would require substantial logistical work to ensure cash reaches those in need with minimal leakage and overhead. But supporters argue the model aligns with current aid approaches and could be piloted in stages before scaling to global coverage. The broader implication for policymakers is clear: the world has the resources to end extreme poverty, but turning that potential into reality hinges on willingness to act.