LIBREVILLE—During a recent talk at the Harvard Kennedy School, students asked me what it would take to turn African countries from consumers of digital and AI-driven technologies into producers. The answer, I told them, starts with heeding the right lessons from the mobile-money revolution.
Digital payment technologies have already transformed Africa. In a single decade, countries that never built a bank-branch network leapfrogged to digital payments. According to the World Bank’s 2025 Global Findex Report, 40% of adults in Sub-Saharan Africa hold a mobile money account, up from 27% three years earlier. As a result, the mobile communications trade association GSMA finds that GDP in Côte d’Ivoire, Ghana, Kenya, Senegal, and Tanzania was 8–10% higher at the end of 2023 than it would have been otherwise.
With policymakers inferring that developing economies can usually bypass the arduous, expensive stages of infrastructure development by adopting technologies built elsewhere, they have applied the same logic to e-government, tax administration, digital identification, and increasingly, AI.
But this approach deserves more scrutiny, especially when applied to AI. Since mobile-money systems generate identity-linked transaction histories and granular economic data, whoever controls them could determine who gains access to the most valuable data and distribution channels for AI services in the world’s fastest-growing population.
This is not to suggest that leapfrogging is a mistake. For capital-constrained countries, adopting proven technology is often rational. But problems arise when imported technologies start substituting for building domestic capacity, and when governments cannot operate, modify, or replace them to meet their own policy objectives—a basic test of digital sovereignty.
Because African countries spend an average of 0.45% of GDP on research and development, compared to a global average of 1.7%, importing ready-made technology looks like common sense. But what started as a pragmatic response to scarcity has hardened into a potentially dangerous development doctrine. By skipping the learning phase, Africans risk foregoing precisely the capabilities that sustain productivity over time.
Development is sequential. It follows from removing binding constraints, not merely circumventing them. Although technology adoption can increase economic activity, it does not necessarily make an economy more productive. Despite GDP growth averaging roughly 4% per year over the past decade, labor productivity in Sub-Saharan Africa remained essentially unchanged in real (inflation-adjusted) terms from 1991 to 2024. Yet total-factor-productivity differences account for more than 66% of the income gap between countries. As AI is adopted, addressing such disparities will become the central challenge.
Many African countries have fallen into the trap of premature automation, adopting AI and digital systems before developing the necessary institutional, infrastructural, and human-capital foundations. As Dani Rodrik of Harvard University has shown, developing countries are reaching peak manufacturing at lower income levels and abandoning it sooner than the original industrializers did.
AI is unlike earlier digital technologies in one crucial respect: it cannot simply be downloaded and deployed. It depends on complementary capabilities such as computing infrastructure, energy, data, technical talent, and firms capable of adapting it locally. The countries that build strategically around these inputs will capture a disproportionate share of the value. Countries that rely on finished systems may still see a productivity bump, but they will not gain the capacity to reproduce it.
The difference between using technology and acquiring technological capability is critical. Knowledge accumulates through the arduous process of building something that doesn’t yet exist. Through trial and error, engineers learn, firms develop capabilities, and regulators gain expertise. This is what Rodrik and his Harvard colleague Ricardo Hausmann call growth through self-discovery. M-Pesa is the clearest African example of this process. Built in Kenya, its engineers and regulators walked away with not just a working product, but deep expertise and institutional confidence—making it a reference point across the continent.
None of this requires reinventing the semiconductor. What matters is whether a country integrates imported technology into a process that it directs, as opposed to installing a complete technology stack that it cannot understand, adapt, or replace. The first leads to greater capabilities, while the second produces only deeper dependencies.
This distinction is becoming more important as technology becomes inseparable from economic and national security. Major powers have already shown that they are willing to restrict access to critical technologies, control semiconductor supply chains, and encroach on others’ digital sovereignty. Governments running essential services on foreign-owned systems have limited leverage when those systems become strategically important. This risk is visible across Africa today. With 360 megawatts of active computing load, the continent accounts for only 0.6% of global data-center capacity, and African governments are increasingly relying on foreign providers for identity, payments, and taxation.
Recognizing that premature automation is both an economic risk and a governance risk, African policymakers will need to distinguish carefully between foreign technologies that can safely be adopted and systems whose ownership or control is critical to state capacity. They also will need to invest more in their own people. Since AI amplifies existing know-how, a country with a deep engineering talent pool and strong institutions can use it to accelerate its domestic innovation, whereas a country lacking these assets risks becoming a permanent consumer of systems designed to serve someone else’s priorities.
The test for determining where a government stands is simple. If it stopped paying a vendor tomorrow, could it keep the service running? If it wanted to switch platforms, do local alternatives exist? And if it obtained the source code, does it have people who could use it? For most African governments, the honest answer to all three is “no”—a shortcoming for which the development community bears responsibility, having long treated digitalization as a service to deliver rather than a capability to build.
Leapfrogging was a rational response to scarcity. But it cannot keep substituting for genuine development. The question is no longer whether Africa should adopt AI (it must, or else consign itself to irrelevance), but which capabilities it must build before, alongside, and because of that adoption.
That is what I told the students at Harvard. African policymakers and development partners will need to think carefully about the next contract they sign, the next procurement decision they make, and the next AI or digital service they deploy. Such decisions are opportunities to recall what lasting development requires.
Copyright: Project Syndicate, 2026.
