A new IMF departmental paper puts numbers on a question Nigerian policymakers have largely addressed in broad terms – what AI could actually be worth to the economy and why its current payoff is so limited.

Titled Unlocking the Potential: Artificial Intelligence in Sub-Saharan Africa, the paper is led by Martin Schindler and Andrew Tiffin and applies a task-based framework similar to one Daron Acemoglu used for the United States.

Over the next decade, AI adoption in Sub-Saharan Africa could increase productivity by 0.2% to 2.1% cumulatively. That tenfold range, the IMF says, is driven almost entirely by policy choices rather than by the technology itself.

The low number is where Nigeria sits today

Under ‘current conditions’ – meaning today’s adoption rates, infrastructure and skills base – the IMF’s model estimates that Sub-Saharan Africa’s productivity would rise by just 0.2% over ten years, translating into a cumulative GDP increase of roughly 0.4%. That is well below the roughly 1% gain projected for Europe and the Western Hemisphere under the same framework. The shortfall is structural rather than technological. Drawing on Microsoft data about AI-related activity on Windows devices, the IMF puts AI diffusion in Sub-Saharan Africa at only 9%. The figures are 30% in North America, 22% in Europe & Central Asia and 21% in the Middle East and North Africa. South Asia, at 8%, is the only region with a lower rate.

Nigeria stands apart in the region. The IMF places most Sub-Saharan African countries in the ‘low exposure’ category because a large portion of their employment involves manual, routine or context-specific tasks that AI cannot easily affect. Only six countries in the region – Nigeria, Botswana, Mauritius, Namibia, the Seychelles and South Africa – have job compositions that more closely resemble those of emerging-market economies than the regional norm. That points to a greater share of roles that AI could support or disrupt. Nigeria is also among the IMF’s top five regional economies for potential productivity gains from AI, along with South Africa, which ranks highest at roughly 0.4 percentage points, as well as Mauritius, Botswana and Namibia.

Why the gains are so small – electricity

The IMF identifies power as the main constraint. In Sub-Saharan Africa, roughly half the population lacks reliable access to electricity. Of the approximately 670 million people worldwide without any access, 85% live in the region. Even so, population growth means the net annual gain is only 5 million people.

Unreliable power comes at a measurable cost. Data from the World Bank Enterprise Survey, cited by the IMF, show that 78% of firms in Sub-Saharan Africa experience routine outages. On average, those outages reduce sales by 8.4%, compared with a 5.2% global average. AI is subject to an even tighter constraint than ordinary businesses – training & operating models rely on centralised, high-performance computing that has to run continuously. A short outage can interrupt a training run, increase latency and result in real losses.

Nigeria’s figures suggest that firms are adapting instead of waiting for the grid. Some 86% of Nigerian firms own or share a generator, the highest proportion among the three countries hosting Africa’s largest concentration of data centres, the figures are 65% in Kenya and 63% in South Africa. Generator power is more expensive and often relies on imported fuel, which the IMF says reduces total factor productivity compared with grid electricity. There is also a fiscal dimension – the IMF estimates that Nigeria’s electricity subsidies amounted to roughly 0.2% of GDP in 2022, while petroleum subsidies contributed another 0.8%. Reliance on self-generated, subsidised power therefore affects public finances as well as private balance sheets.

A small footprint, and an opportunity

Sub-Saharan Africa has roughly 0.4 gigawatts of data-centre capacity relevant to AI, which is under 1% of the global total. Figures from datacentermap.com show 62 facilities in South Africa, 25 in Nigeria, 19 in Kenya, and single-digit counts in other countries, according to the IMF.

The IMF models a scenario in which the region holds just 0.5% of global AI data-centre capacity by 2035 – a modest ambition. Yet meeting even that target would add electricity demand equivalent to roughly 10% of the region’s total installed generation capacity in 2023. A small increase in global market share, in other words, would still amount to a double-digit percentage of what the region currently produces.

This is the same tension seen from another angle in Nigeria’s CBN-driven data centre boom. The CBN’s localisation directive is attracting investment for financial-sector reasons. The IMF paper, though, points to a broader source of demand – as AI expands beyond banking, it could put pressure on the grid once it moves beyond its current tiny base. The paper presents that prospect as both an opportunity and a risk.

Data centres require substantial, predictable power over the long term, so they can serve as anchor tenants and make grid expansion and renewable projects ‘bankable’ – something dispersed household demand often cannot do. Kenya offers the clearest example – Microsoft and G42 have committed $1 billion to a 100-megawatt green data centre campus powered by geothermal energy, in a country where renewables already account for more than 90% of generation. Elsewhere, Cassava Technologies and NVIDIA agreed to a $700 million deal to deploy 12,000 GPUs in South Africa, Nigeria, Kenya, Egypt and Morocco. The IFC has also committed $100 million to expand Raxio Group’s data centres in Ethiopia, Angola, Mozambique and Côte d’Ivoire.

The IMF cautions that Nigeria’s broader gains will depend on the design of its power purchase agreements and grid-extension mandates. Rising demand could bring genuinely new capacity online, but it could also leave households and other businesses competing for access to a grid already under strain.

Skills are a second binding constraint

Beyond electricity, the IMF identifies a severe gap in computing access as a force behind brain drain – just 5% of Africa’s AI talent can reliably access the high-performance computing needed for serious research. That shortage creates an ‘iteration gap’. A G7 practitioner might retrain and test a model every 30 minutes, whereas an African counterpart could wait up to six days for one run. It is easy to see why skilled workers move toward markets with plentiful compute. Nigeria’s Three Million Technical Talent programme, launched in late 2023, is a genuine attempt to narrow the divide. Still, the IMF makes clear that training people more quickly will not, on its own, keep talent from leaving unless local computing capacity grows as well.

The policy takeaway

The IMF’s 0.2% estimate is not a forecast of AI’s technological potential. It reflects today’s constraints – power, connectivity, skills, capital and regulation. If enough of those constraints are addressed quickly, the same model puts the gain at more than ten times that level – 2.1%, alongside an increase in GDP growth of roughly half a percentage point a year. For Nigeria, then, the current wave of data-centre investment, whether driven by the CBN or not, is necessary but not sufficient. Rack space and fibre by themselves will not deliver the upper end of the IMF’s range. That will depend on whether the grid, skills pipeline and regulatory institutions expand quickly enough to support the infrastructure being built on top of them.

Two sides of the same infrastructure question

This is really the second half of a story already unfolding in Nigeria’s financial sector. Industry figures view the CBN’s June 2026 directive, which requires transaction data to be stored locally by January 2027, as an initial move toward broader data sovereignty. They expect the policy to extend to oil and gas, government, identity and immigration systems, education, and healthcare. Billions are already being invested in Nigerian data-centre capacity. Operators ranging from OADC and Rack Centre to Equinix say the pipeline is real.

The IMF paper arrives at the same tension by a different route. The CBN story is about compliance deadlines and forex savings. The IMF, though, raises a more basic question – even if Nigeria builds every data centre required by the CBN’s localisation push, will the power grid, skills base and regulatory institutions be strong enough to turn that infrastructure into a genuine AI productivity dividend? Or will it simply create costly server space running on diesel?

The CBN has given Nigeria a reason to build data centres. Whether that investment delivers the gains described by the IMF will depend on decisions about power, skills and regulation – decisions that have barely begun.

  • Henry Nduka Onyiah works as a Cyber & Technology Risk Advisor and serves as an Independent Non-Executive Director at a Nigerian Financial Institution. He writes in his personal capacity, and the views expressed are entirely his own.