You can listen to this podcast on Spotify, Apple Podcasts, or wherever else you get your podcasts. Or you can watch this conversation on YouTube.
If you’ve been following the buzz about artificial intelligence in Africa, you’ve probably heard a mixture of grand optimism and stark warnings. Some argue the continent could leapfrog straight into an AI-powered future, while others warn that AI will fall at the first hurdle on the continent.
So, can AI take off in Africa? To talk us through this question, in this episode of Ideas in Development, we are joined by Rose Mutiso, founder of the African Tech Futures Lab and former Research Director at the Energy Growth Hub, to explore what an AI boom would actually require in African countries, and what needs to happen before the hype can translate into widespread, locally useful adoption.
AI in Africa: The basics
Popular narratives about AI often skip over the physical and economic foundations that make digital systems work. In high‑income countries, those foundations often feel like a given, with near total access to electricity, constant connectivity, and cloud services which are reliably a click away. But when those inputs are scarce, expensive, or unreliable, like in many parts of Africa, AI doesn’t arrive simply as a frictionless ‘app’, since it places a set of additional demands on already-stretched infrastructure.
So the answer to whether AI can take off is in some ways a basic one. To enable African countries to capitalise on the AI era, countries will need the enabling conditions that make modern digital economies function at scale.
Three keys: Power, connectivity, and data
So what infrastructure is essential for AI to become widespread? Rose boils it down to three fundamentals: power, connectivity, and data.
Power. AI is more than just software; it is computation, and computation consumes energy, especially in data centres (Rose calls these the physical backbone of the AI economy).
“One is power. AI is energy hungry.”
Connectivity. It’s not enough for people to have occasional mobile access. AI pushes requirements upwards, in terms of bandwidth, reliability, and the backbone infrastructure that supports modern digital services.
“In the AI age, it’s not just access, it’s really connectivity infrastructure that is needed.”
Data. AI systems need data. When datasets don’t exist, or are not open, structured, and relevant to local realities, countries will only be passive consumers of tools built elsewhere, rather than shaping systems that work for them.
“In poor countries like (some of those in) Africa, where your data is scarce, you’re just invisible. You’re literally invisible in the AI ecosystem.”
Data centres: The physical backbone of the AI economy
At its most simple, a data centre is a warehouse that’s filled with computers, or servers, that store data and process information for websites, cloud services, and AI.
“People often think the internet is invisible, but it’s actually quite physical.”
Rose emphasises that the internet is a physical system comprising of buildings, servers, cables, energy supply, cooling, and maintenance.
And data centres are the infrastructure layer that makes cloud services work. Your phone or laptop might feel like the site of computation, but much of the storage and processing happens somewhere else. That’s why location and capacity are important, since the geographic distribution of data centres affects costs, latency, resilience, and who controls digital infrastructure.
In the AI era, the physical nature of the internet becomes more consequential. AI raises the volume, speed and intensity of computing demand, which makes data centres more important, and their energy use becomes a central part of the story. If Africa wants a stronger stake in the AI economy, building digital infrastructure (including data centres) is part of the equation, as is building the energy systems that make them viable.
How far behind is Africa on AI infrastructure?
The gap with the rest of the world is large. On electricity, Rose points to the sheer magnitude of unmet need and the compounding problems of reliability and affordability, even where grid connections do exist.
She notes that Africa’s electricity prices are among the highest in the world and that power supply is often unreliable, two issues that are especially damaging for data centres and digital businesses that require consistent uptime.
And on data centres, she describes a global system that remains largely absent on the continent. Rose highlights that Africa has about 166 times the population of Switzerland, but only double the amount of data centres. Infrastructure has scaled very unevenly across different regions of the world, and that shapes who benefits from technological shifts.
But this worrying continent-wide story does cover up some important bright spots. Rose mentions South Africa, Kenya, Nigeria, Morocco, and Egypt as promising areas where fibre-optic networks, data centres, and digital services are expanding, and where research capacity (including supercomputing clusters) is emerging.
And Rose emphasises that the continent does not need to ‘finish’ electrification or connectivity everywhere before it can use AI at all. Instead, she argues for sequencing: starting with the places and sectors where adoption is feasible now, while building the fundamentals for broader access over time.
“You can sequence the usage and the growth… you don’t have to have the perfect… total system in place.”
Can Africa leapfrog into AI, like it did with mobile?
The conversation then takes one of the more widespread AI and development tropes: leapfrogging. Mobile phones bypassed the need for landlines in Africa, whilst mobile money helped reach people excluded from traditional banks. Many argue that AI could do something similar.
Rose is cautious of this framing, warning against assumptions that AI will be a similar kind of leapfrog technology as earlier digital tools.
“AI is not a classic leapfrog technology.”
Classic leapfrogging works when a new technology bypasses legacy infrastructure. AI, however, depends on heavy infrastructure, including energy, compute, and data systems.
That doesn’t mean Africa can’t innovate on the margins. Rose notes that Africans already access tools like ChatGPT by using infrastructure hosted elsewhere: including data centres in Europe, undersea cables, and global cloud platforms. But she argues that relying on offshore infrastructure is not a sustainable long‑term foundation for inclusive, locally governed AI ecosystems.
“We are able to access infrastructure offshore elsewhere, but this is not a long-term sustainable proposition.”
Africa can absolutely participate in AI already, and there may be smart workarounds (including regional sharing of infrastructure). But scaling AI in a way that builds local capability and avoids dependency will require investment in the underlying ecosystem.
Three steps to unlock AI’s potential in Africa
Rose lays out a practical agenda for AI in Africa, much of which should already be part of development strategies on the continent.
1) Build affordable, reliable power at scale
“We just need to build more power, full stop.”
That means investing across the electricity value chain, generation, transmission, and distribution, to expand access and improve reliability. This is needed for AI, but is also badly needed more broadly to support households, industry, healthcare, education, and more.
This agenda includes a potential win‑win: data centres could become anchor customers for utilities. Because they demand high-quality, reliable power and consume large, predictable loads, they can provide steady revenue and justify grid upgrades, if managed well. Rose contrasts this with rich countries where AI-driven power demand is often framed as a threat to the grid; in contexts where demand growth is needed, it could be an opportunity, but only with careful planning and coordination.
2) Governments as enablers, not planners
Rose’s second step is about the state’s role. She critiques the tendency to produce sprawling strategy documents that promise everything and deliver little.
Instead, she proposed that governments focus on the specific levers they can pull effectively. In practice, that includes:
Clear, predictable regulation (for power contracts, land, data, competition) that creates confidence for investors.
Reducing friction in permitting and deployment so infrastructure projects can actually happen.
Harmonising regional standards, especially around data governance, to avoid a patchwork of rules that fragments markets.
Creating demand by digitising public services, and in some cases requiring sensitive government data to be hosted locally, which can catalyse domestic data centre ecosystems.
Targeted public investment rather than trying to fund everything: prioritising key sectors, improving public data systems, and supporting skills through education.
Governments have an important role to play as enablers, without pretending they can centrally plan an AI transformation.
3) Reduce barriers to entry for local researchers and entrepreneurs
Rose’s third step shifts from infrastructure to the conditions facing local AI innovators. She argues that it’s important to think about adoption in stages: new technologies typically diffuse through early adopters first, particularly firms.
Then she lists the practical blockers: high costs, limited access, weak data, and thin research networks.
“I think the main barriers are, prohibitively expensive cloud computing and compute access,”
She points out that major cloud providers have limited regional coverage (with many services concentrated in South Africa), and that subscription costs and payment frictions can make access hard for African researchers and institutions.
Solutions include subsidised access programmes, equipment donations and partnerships, stronger public data systems, and more support for experimentation, including very small grants that can materially change what is possible.
“small grants can have a huge, huge difference.”
Lowering barriers for these groups in the short term doesn’t require billion‑dollar infrastructure projects.
An African AI conversation led by Africans
We ended the episode discussing Rose’s new initiative, the African Tech Futures Lab, which aims to foster deeper, more grounded thinking about AI from African perspectives.
“we are going to be on the lookout for rising … thinkers, African thinkers, both in Africa and the diaspora, on AI.”
There are a growing number of Substacks about the economics of technology and AI. There are very few by Africans, about Africa. Rose wants to change that, and cultivate a cohort of voices that can help to cut through the buzz, translate complexity for policymakers and the public, and articulate what AI means in African contexts. For more details, check out the African Tech Futures Lab website.
Read more from Rose Mutiso
This episode drew on a number of excellent articles written by Rose over the past few years, you should check them out, along with her excellent Substack Kibao.
This is a great episode and it really highlights some interesting perspectives. For instance, I didn’t know about the disparity in the concentration of data centres in North Africa compared to South Africa and Switzerland as a whole. It also shed some light on new ways of thinking about accessibility challenges for researchers to cloud services especially in the context of the technological leap into AI.
Credit to the VoxDev team and to Rose for such an insightful episode.