Every decade, a new technology arrives in Africa carrying the same fundamental question beneath its promise: who will own the infrastructure, and who will own the returns? The telegraph, the telephone, the internet- each was marketed as connectivity, and each delivered connectivity on terms that kept the value-capture architecture in the hands of those who built and controlled the networks. Artificial intelligence is arriving with greater fanfare and greater consequence than anything that preceded it, and the question is the same.
The healthcare applications are genuinely promising. Nigeria’s doctor-to-patient ratio, approximately 1 to 4,000, means that AI-assisted diagnostic tools represent not just technological improvement but medical access at scale. Rwanda’s drone-delivered blood supply, operated through a partnership with Zipline, demonstrates that AI-powered logistics can bridge the infrastructure gaps that ground-based systems cannot cross. Agricultural prediction tools that help smallholder farmers navigate rainfall variability, pest cycles, and market timing are not conveniences; they are survival infrastructure. These applications are real, and their potential to address the structural service deficits that African governments have failed to close through conventional means is significant.
But the same tools that provide healthcare access, agricultural guidance, and financial services generate data, and the ownership of that data, the infrastructure through which it is processed, and the algorithmic systems that derive value from it are not, for the most part, African. The continent that provides the use cases, the market scale, and the human data inputs for AI applications developed in Silicon Valley, Shenzhen, and Bangalore is not, in most cases, capturing proportional returns from those applications. This is the extraction logic applied to intelligence rather than minerals, and it is more difficult to see precisely because the extraction is invisible; it happens at the level of data flows, model training, and algorithmic optimisation that leaves no physical trace in the environment where the value was generated.
The economic displacement risk compounds this. The World Economic Forum’s projections, later updated and revised but consistently directional, identified Africa’s labour-intensive manufacturing and agricultural sectors as among those most exposed to automation. Ethiopia’s textile manufacturing, which was positioned in the 2010s as a potential driver of export-led industrialisation on the East Asian model, faces an automation wave from the very supply chains it was building to supply. Botswana’s retail sector is already experiencing the substitution of human agency roles with AI and digital systems. For a continent where over eighty per cent of employment is generated by small and medium enterprises operating in labour-intensive sectors, the automation timeline is not an abstract policy question. It is an immediate economic risk.
The path to a different outcome is not primarily technological. It is institutional. Africa needs data governance frameworks that assert national and continental sovereignty over the data generated by African users and African economic activity, frameworks equivalent in ambition to the European Union’s GDPR but designed for African economic conditions and African interests. It needs investment in the computational infrastructure, data centres, connectivity, and power supply that makes domestic AI development viable rather than forcing reliance on cloud infrastructure owned and operated by companies whose primary accountability is to shareholders in other jurisdictions. It needs educational systems that produce not just AI consumers but AI architects, people who can design, train, evaluate, and govern the systems that will increasingly shape economic life.
Rwanda, Ghana, and Kenya are demonstrating that African governments can move meaningfully in this direction, through national AI strategies, public-private investment in digital infrastructure, and educational curricula that take computational literacy seriously from early ages. These are not sufficient models for the continental challenge, but they are evidence that the institutional capacity to respond to AI’s arrival on African terms is not absent. It is unevenly distributed and under-resourced relative to the scale of the challenge.
The historical analogy that should be most disturbing for African policymakers is not the internet’s arrival in the 1990s. It is the agricultural economy’s transformation in the 1970s, when Nigeria and other commodity-export economies allowed an external windfall, oil revenue, to displace domestic productive capacity without building the institutional infrastructure that would have made the transition generative rather than destructive. The groundnut pyramids of Kano disappeared not because oil was more valuable than agriculture, but because no one built the institutional bridges that would have allowed both to coexist and compound. AI presents the same risk at a different scale: a technological windfall that displaces existing economic activity without generating equivalent domestic value if the institutional architecture to capture that value is not built in parallel.
Africa will not solve its AI challenge by watching how others solve theirs and then adopting the solutions. The continent’s conditions, its data geography, its infrastructure constraints, its demographic profile, its economic structure, require AI strategies designed for Africa. That requires African institutions, African researchers, African policymakers, and African capital. All of those exist. What has been missing, repeatedly, is the political will to direct them toward the long-term structural work rather than the short-term adoption of solutions built elsewhere. The AI moment is still early enough for Africa to enter it as an architect rather than a market. That window will not remain open indefinitely.
