From health diagnostics in Nairobi to smart farming tools in Accra, artificial intelligence is already part of everyday work across the continent. Whatever shape the Africa AI innovation and technology development future takes, it will be settled by decisions made in the next few years, not the next few decades. The Carnegie Endowment for International Peace estimates that AI could add between $2.9 billion and $4.8 billion to Africa’s economy by 2030. But money flowing through a continent is not the same as value staying in it. The harder question is whether Africans will build the systems that shape their economies, or simply rent them from somebody else.
A Continent Full of Strategies
Nobody can accuse African governments of being quiet on this. Ethiopia, Nigeria, Libya, Zambia and Mauritania published national AI frameworks in 2024. Côte d’Ivoire, Kenya, Namibia, Lesotho and Tanzania followed in 2025. At least fifteen countries now have a plan of some kind, and the African Union released its own continental strategy in 2024.
The momentum has carried into 2026. African Union ministers met in Tangier, Morocco, in April, and 49 countries plus the AU have endorsed the Africa Declaration on Artificial Intelligence, which calls for more investment in local infrastructure, talent and innovation. Ghana’s strategy, launched the same month, goes as far as calling AI a “sovereign capability.”
That is real ambition. But writing a strategy is the easy part. Delivering it is where most countries start to struggle.
Where the Dream Meets the Cables
Africa is home to roughly 18 per cent of the world’s people, yet it holds less than one per cent of global data centre capacity. McKinsey found that the continent’s five biggest data centre markets combined have less capacity than France on its own. Only about three per cent of the world’s AI talent is based in Africa. And anyone who has sat through load-shedding in Johannesburg knows that unreliable electricity makes every expansion plan harder.
This matters because you cannot train, host or properly govern AI without computing power. A country that depends on someone else’s servers also ends up depending on someone else’s terms.
The Kenya Test
Few projects show the tension better than the proposed $1 billion data centre in Kenya involving Microsoft and the Emirati firm G42. President William Ruto drew attention to how much electricity a facility of that size would demand, and he made clear that new power generation would be needed. Kenyan officials say talks are still ongoing, so the final shape of the deal is not yet clear.
Whatever happens, the episode captures the trade-off every government faces. Investment is badly needed, but so are cheap, reliable energy and clear limits on long-term dependence. Researcher Sanusha Naidu of the Institute for Global Dialogue has pointed out that data centres are also heavy users of water, which puts pressure on local communities. She compared the situation to the textile investments of the 1990s, when host countries often subsidised the companies that came in. With data centres, she argues, the burden can be even heavier.
Who Really Holds the Cards?
Here is the encouraging part. Geopolitical analyst Priyal Singh of Signal Risk argues that the global AI industry is so fragmented and competitive that African states actually have more room to negotiate than people assume. He points to the pushback over Starlink’s expansion in parts of Africa as an example of governments becoming bolder with global technology firms.
It also helps that the suitors come from everywhere: the United States, Europe, the Gulf and China. Naidu’s advice is practical. Whether the investor is American, European or Chinese, a government should ask the same question: what do we get back, in jobs, skills and lasting capacity?
Some deals already hint at a better model. The $720 million “Africa AI Factory” from Cassava Technologies and NVIDIA mixes international capital with African leadership. It is not perfect, but it shows that partnerships do not have to mean surrender.
Built in Africa, for Africa
Imported systems often miss the realities on the ground. At a ministerial dialogue held alongside the 2025 Global AI Summit, African leaders stressed that the continent’s language diversity, energy limits and patchy connectivity call for solutions designed for those conditions. That means lightweight, energy-efficient models that can run where bandwidth is thin and power is unpredictable.
It also means data. Building datasets in indigenous and local languages is not just a cultural nice-to-have. It is how a farmer in rural Tanzania or a trader in Kano gets a tool that actually understands them, and it reduces reliance on foreign models trained on someone else’s world.
Talent is the other half. Togo has promised to train 50,000 people a year in AI skills. If more countries match that kind of target, the brain drain that has long hollowed out African tech could finally slow down.
The Missing Voice: Ordinary Citizens
Most of this debate happens in conference halls and ministry offices. Joseph Asunka, who leads Afrobarometer, warns that deals negotiated at elite level and then handed down to citizens create a trust gap. If people do not trust what their governments are agreeing to, they are less likely to embrace fintech, e-commerce and digital government services, the very tools that are meant to benefit them.
Governments also need to get their own house in order. South Africa pulled a draft national AI policy earlier this year after officials found references that could not be verified and appeared to have been produced by AI tools. It was an embarrassing lesson: regulating this technology demands care, expertise and human checking.
So, Creator or Consumer?
The honest answer is that Africa will be both, and the goal should be to tip the balance. No country on earth is fully self-sufficient in AI, and Africa will keep relying on global supply chains, investment and expertise. The real fight is over terms: who owns the infrastructure, who controls the data, and who captures the profits.
There are practical steps that would help. Regional data centres powered by clean energy could serve several countries at once. Pooled funds and blended finance could give homegrown startups steady backing. Diaspora professionals could be encouraged to come home and build. Initiatives like the UNDP’s timbuktoo programme, which supports innovation and entrepreneurship, are still small next to global AI spending, but they point in the right direction.
The independence generation argued that political freedom meant little without control over economic resources. Today’s version of that argument is about data, computing power and code. If Africa treats AI as a field to shape rather than a product to buy, the continent can move from being a market to being an author. If it does not, the future will be written elsewhere and sold back to it.
