World Bank's 2026 development report argues AI can give developing economies a leapfrog — if they build the basics first
The Bank estimates 4.5% of jobs in low- and middle-income countries are at automation risk versus 14.2% in high-income ones, and says the leap depends on power, connectivity and skills being built first.
On August 4, the World Bank released its World Development Report 2026, titled The Promise of Artificial Intelligence, and the headline finding cuts against a reflex that has shaped most public AI discourse in rich countries. The Bank estimates that 4.5% of jobs in low- and middle-income countries are at risk of automation from AI, compared with 14.2% in high-income countries, while 16.2% of jobs in developing economies could see their productivity meaningfully boosted by the technology. The asymmetry is not subtle, and it is the reason Chief Economist Indermit Gill told reporters that developing countries “do not need large models or big data centers to reap its benefits” — but that they also “must hurry,” according to the Bank’s own press release.
The framing matters because the conventional AI race — build the biggest model, train it on the most tokens, deploy it from a hyperscale data center — is one that almost no developing economy can win on the terms rich countries have set. The Bank’s argument is closer to the opposite: the productive use cases available to a country like Nigeria, Vietnam, or Kenya do not require frontier-scale compute, and pretending otherwise is a way of letting the infrastructure gap become an excuse for inaction. As Nairametrics reported on the same day, Gaurav Nayyar, one of the report’s lead authors, framed the moment as a “narrow window” — implying that the conditions for adopting AI cheaply and productively will not stay open indefinitely as the underlying technology and its global concentration harden.
Nigeria is one of the countries the report holds up as evidence that adoption is already happening at speed. AI uptake among formal businesses in Nigeria has accelerated sharply over the past three years, a shift visible enough that the World Bank flags it as a marker of how quickly a developing economy can move once the basic enablers are in place. The point is not that Nigeria has solved the underlying structural problem; it is that the rate of change inside the productive sector is faster than the pessimist case would suggest, and faster than the infrastructure conversation usually acknowledges.
That last clause is where the report’s policy weight sits. The headline promise — a decade-to-century leapfrog for developing economies, in the Bank’s framing — is conditional on three things that the Bank argues have to be built or substantially expanded first: electricity generation, broadband connectivity, and a workforce with the skills to use the tools productively. Without those, even cheap, narrowly targeted AI applications cannot deploy at the scale that would generate the productivity gains the 16.2% figure implies. The report is, in effect, a warning that the gap between “AI is available” and “AI is being used productively in an economy” is going to be widened by infrastructure deficits that are entirely solvable, but only with the kind of capital and policy attention that developing economies are currently not receiving at the required rate.
The broader context is also why the Bank chose to publish this now. Global growth has been weak for several years, the productivity slowdown that pre-dated the AI wave is still unresolved, and developing economies have less fiscal room than they did in earlier technology transitions to subsidize the diffusion of a new general-purpose technology on their own. The Bank’s case is that AI is unusually well-suited to a development moment precisely because the cheapest applications are not the most compute-intensive ones — but that this only helps if the electric grid, the fiber, and the trained workforce are already there when the model arrives. Building all three at once is the actual project. “Big data centers,” as Gill’s quote frames it, are not.
What the report is implicitly pushing back against is a development-finance discourse that has, until recently, treated AI as either a frontier-labor-market problem for rich economies or a sovereign-infrastructure problem for the few countries that might host hyperscale compute. The Bank is trying to redirect attention toward the much larger group of countries where AI could plausibly raise productivity in agriculture, formal-sector services, logistics, and customer-facing operations within a few years — provided the basics are built. Whether that redirection lands at the level of actual concessional lending, regulatory technical assistance, and skills-program funding is a separate question, and one the report does not pretend to settle. What it does is put a number on the stakes, attach a deadline to the window, and tell developing-economy governments that the goal is not to mimic Silicon Valley but to do the unseeming groundwork that makes the technology actually work.
Sources
- AI Offers Lifeline to Developing Economies in an Era of Weak Growth — World Bank
- World Bank tells developing economies to stop chasing big AI models — Nairametrics