The global AI economy is usually discussed as a contest among a few major powers, measured in model releases, chip supply and national strategy documents. That framing has little room for a question that matters more on this continent: can Africa participate as a builder of value, or only as a market for systems designed elsewhere? The answer will be settled less by the arrival of new tools than by who owns the layers beneath them.
The investment data sets out the starting position plainly. UN Trade and Development reported that Africa attracted just 3% of total data-centre investment in 2024, and recorded only 18 fintech projects that year against 206 in developing Asia. The stock position is thinner still: the continent accounts for less than 1% of global data-centre capacity, and sub-Saharan Africa has roughly 0.1 data centres per million people against a world average of 0.5. Nor is this simply a Global North and Global South divide. Around 80% of greenfield digital projects going to developing countries over the past five years landed in just ten of them, mostly in Asia.
What makes those figures more urgent is how fast the denominator is now growing. UNCTAD's preliminary 2026 data shows data centres captured more than a fifth of global greenfield project value in 2025, with announced foreign direct investment above $270 billion, concentrated in a handful of hosts led by France, the United States and the Republic of Korea. Only a few emerging markets appeared among the top recipients. A small share of a small market is a manageable gap. A small share of the fastest-growing category of global investment compounds into something harder to close.
This matters because AI is not primarily an applications question. It depends on computing capacity, reliable power, connectivity, locally relevant data, and people able to build and govern the systems. Each of those is a binding constraint, and they interact. Power is the least discussed and among the most severe: frontier model training draws tens to hundreds of megawatts, a load most national grids on the continent cannot supply reliably, which in turn deters the data-centre investment that would justify grid expansion. The global concentration is stark by comparison, with the United States holding about a third of the world's top supercomputers, over half its computing power, and roughly 70% of global private AI investment.
Data is the second constraint, and here the gap is cultural as much as technical. Africa is home to more than 2,000 languages, yet under 1% of widely used training data reflects African linguistic contexts. The consequence is not only poor translation. It is systems that misread accents, cannot process indigenous clinical terminology, and fail quietly in exactly the public-service settings where they are most often proposed as a solution. When a language is absent from the data, its speakers are absent from the product, and no amount of downstream deployment corrects for that.
Governance is the third, and it is the one most easily ceded by default. Of the seven major international AI governance initiatives active by 2024, only the G7 countries participated in all of them, while 118 nations took part in none. Most of those are developing countries. Rules on data, liability, safety and market conduct are being written now, and they will apply to African markets whether or not African institutions helped shape them.
None of this argues for pessimism, and the ground is moving. Cassava Technologies has opened an AI factory in South Africa, renting GPU capacity to African developers and businesses as part of a partnership with NVIDIA to deploy 12,000 GPUs. IXAfrica's Nairobi campus is operational and has partnered with Safaricom on AI-ready infrastructure. Six of the continent's largest operators, including MTN, Airtel, Orange, Vodacom, Ethio Telecom and Axian, have joined a GSMA-led effort to build language models in African languages, framed as work done in Africa, by Africa, for Africa. The World Economic Forum has estimated that resolving the compute constraint could unlock $1.2 to $1.5 trillion in economic value by 2030. Policy intent exists too: 86% of developing countries now have national digital strategies, including 80% of least developed countries, up from under half in 2017.
The risk in that list is reading it as arrival rather than as a beginning. Deployment without local capacity deepens dependence, because using systems hosted and governed elsewhere means paying rents abroad, accepting terms set abroad, and losing the learning that comes from building. A country can raise AI adoption sharply while its position in the AI economy weakens, and adoption statistics will not show the difference.
So Africa's place will not be determined by a single continental ranking, and treating readiness indices as a scoreboard obscures more than it reveals. It will be shaped by specific and contestable choices: whether power generation is treated as digital infrastructure, whether public data is made usable and protected at the same time, whether language datasets are built as shared public goods rather than proprietary assets, whether procurement rules favour local capability, and whether universities and firms can retain the people they train. Those choices sit with governments, universities, businesses and communities, and they will be made differently in Lagos, Nairobi, Kigali and Cape Town.
The goal is not to reproduce another region's path, which was built on conditions that no longer exist and priorities that were never African. It is to build the conditions for innovation that reflects African needs and expands African agency, which means treating infrastructure, data, skills and institutions as the substance of the opportunity rather than as prerequisites to be cleared before the interesting work begins.
