Two decades ago, a digital revolution transformed the lives and livelihoods of millions in sub-Saharan Africa as mobile phones swept across the continent. A region where a small fraction of the population had landline phones or bank accounts was able to leapfrog old telephone monopolies and infrastructure. Mobile phones enabled farmers to access advice on pests and weather, and the unbanked to send and receive mobile money. Now, artificial intelligence (AI) promises to be an even more potent force for boosting productivity and helping poorer countries overcome shortages of skilled teachers and doctors.
However, while opportunity beckons, Africa is being left behind. PWC, a professional-services firm, estimates that AI could add nearly $16 trillion to global economic output by around 2030 (compared to 2017). McKinsey, a consulting firm, independently arrived at a similar figure, but now predicts this could rise by another 15-40% due to newer forms of AI such as large language models. Yet, Africa, home to roughly 17% of the world’s population, is projected to receive a boost of only $400 million from AI in its annual GDP by 2030, representing a mere 2.5% of the total. This is primarily attributed to its lack of digital infrastructure.
As a result, instead of helping to narrow the productivity and income gap between Africa and richer countries, AI seems poised to widen it. Take Nigeria, a regional tech hub, where the average wired internet download speed is one-tenth of Denmark’s. The majority of broadband users in Africa’s most populous nation are limited to mobile internet, which is even slower. A growing number of underwater cables are connecting the continent to the wider world, with more on the horizon. These include Meta’s 2Africa, the world’s longest undersea connection. But a shortage of onshore lines to carry data inland will leave much of that capacity underutilized.
In some ways, Africa’s weak digital infrastructure can be explained by the success of its mobile revolution, where privately owned telcos entered newly liberalized markets, disrupting and displacing incumbent operators. These not-so-new firms are still experiencing rapid growth, with the top 15 averaging a 29% revenue increase over the past five years. However, their jump over landlines is now coming back to bite them. In much of the developed world, the basic infrastructure of telephones—junction boxes, telephone poles, or underground cable conduits—has been repurposed to provide fast fiber-optic broadband. Yet, Africa is often starting from scratch.
The lack of connectivity is compounded by a shortage of the heavy-duty data centers necessary to process the vast amounts of data required to train large language models and run AI-powered applications that could boost Africa’s economic growth. Today, much of the content and processing needed to keep websites and programs running is stored in the cloud, which is comprised of thousands of processors in physical data centers. However, Africa has far fewer of these compared to any other major continent. Without nearby data centers, bits and bytes have to undertake long round trips to centers in cities like Marseille or Amsterdam for processing, resulting in lagging applications and hindering efforts to stream high-definition films.
The closer data is to users, the faster it can reach them: films can zip across to viewers from one of Netflix’s African servers more quickly than you can say “Bridgerton.” The more cable landings and local data centers there are on the continent, the more resilient its network becomes, especially if undersea cables are damaged, as occurred earlier this year when internet access was disrupted across much of West Africa. All these new data centers will require more energy as they expand. AI, which involves complex calculations that demand even more computing power, will further increase demand. A rack of servers needed for AI can consume up to 14 times more electricity than a rack of normal servers. They also require industrial air conditioning, which guzzles massive amounts of power and water, particularly in increasingly hot climates. However, Africa is so short of electricity that approximately 600 million of its people lack power. In Nigeria, which suffers 4,600 hours of blackouts annually, data centers are forced to rely on their own natural gas-powered generating plants to keep the lights on and the servers humming. Although many centers across the continent are turning to renewables, wind and solar are too erratic to provide continuous power.
Edge computing, where more data is processed on the user’s device, is promoted as a way to bring AI-powered technology to more Africans. But it relies on the presence of numerous smaller and less energy-efficient data centers, and on users having smartphones powerful enough to handle the calculations. While roughly half of mobile phones in Africa are now smartphones, most are cheap devices lacking the processing power for edge computing. In 18 of the 41 African countries surveyed by the International Telecommunication Union, a minimal mobile-data package costs more than 5% of average incomes, rendering it unaffordable for many. This might explain why nearly six in ten Africans lack a mobile phone and why it is not profitable for telcos to build phone towers in many rural areas. “Approximately 60% of our population, representing about 560 million people, have access to a 4G or a 3G signal next to their doorstep, and they’ve never gone online,” says Angela Wamola of GSMA, an advocacy group for mobile operators. Every subsequent yet-to-be-connected African is more expensive to reach than the last and brings fewer returns as well. And new phone towers in remote areas, which typically cost $150,000 each, still require costly cables to “backhaul” data.
Part of the solution to Africa’s connectivity problem may lie in partnerships between mobile-phone operators and development institutions. Existing telcos have knowledge of the terrain and the political landscape that can make laying cables a delicate task. International tech firms such as Google or Microsoft are well-positioned to assume more risk by laying their own cables and building data centers. Equipment providers and other multinational companies can fill skill gaps. China’s Huawei, for example, is constructing 70% of Africa’s 4G networks. Startups employing cheaper technologies are exploring ways to help remote communities get connected. Africa’s connectivity mix will likely be as diverse as its people, encompassing everything from satellites that can be deployed by companies like Starlink to reach rural areas, to improved 4G networks in medium-sized cities.
Some foreign firms are investing in data centers in Kenya and Nigeria, but not enough of them. There is also some experimentation with how to power them. Kenya’s Ecocloud Data Centre, for instance, will be the continent’s first to be fully powered by geothermal energy, a stable source of renewable power. Since Kenya’s grid has ample green energy available, it is an attractive location to build more data centers. But given how many power sources your correspondent switched between to write this article, and how many patchy internet connections interrupted her work, much still needs to be done to improve infrastructure. This is even more crucial if Africa’s animators, weather forecasters, quantum physicists, and computer scientists are to realize their potential. Even small-scale farming, which sustains more than half the continent’s people, stands to benefit from enhanced access to AI.
Frustratingly, the case for improving Africa’s digital infrastructure is not new. “Gosh! I can’t believe, 15 years later, we’re still having this conversation,” says Funke Opeke, whose company, MainOne, built Nigeria’s first privately owned submarine cable in 2010. Unless significant investments are made soon, the same conversation may be taking place another 15 years from now.
