African demand for computing infrastructure must rise from 0.4 gigawatts today to 1.5 to 2.2 gigawatts by 2030, according to McKinsey, but the continent has not yet stabilized the power grids that supply its households. The IMF estimated in 2026 that AI’s gain for African economies caps at 0.2% of GDP under current conditions, limited by a set of constraints including electricity, digital infrastructure, skills, and institutions.

The Essential Points

  • Electrical demand from African data centers must triple by 2030, on a network already insufficient for households.
  • Kenya considered safeguards on Microsoft-G42 projects, raising the question of their grip on national capacity (source: Semafor, May 6, 2026).
  • The IMF values AI gains for Africa at 0.2% of GDP under current conditions, versus roughly 1% in Europe and the Western Hemisphere.
  • The trade-off is institutional: without public decision to consolidate networks and treat computing as a public good, AI risks amplifying external dependence.
  • Pathways exist—submarine cables, regional networks, public-private financing—but their rollout remains uneven and fragmented.

Kenya Asked the Question No One Wanted to Ask

Several data center projects are being considered in Kenya. The question of their electrical feasibility remains central: Kenyan networks risk suffering from large-scale digital infrastructure projects deployed without first ensuring reliable household power supply.

Kenya has an installed capacity of roughly 3.3 gigawatts, but service reliability is limited by losses, technical failures, and network constraints. A large-scale data center would absorb a significant share of this capacity, risking amplified power cuts already affecting millions of households. This dilemma reflects an infrastructure constraint, not a rejection of AI.

The Kenyan dilemma illustrates a major obstacle to AI deployment in Africa. The continent has developers, start-ups, and ambition, but lacks kilowatts. This shortage results from decades of underinvestment in basic infrastructure, worsened by financing models that long favored quick-return projects on distribution networks.

0.4 Gigawatts for a Continent of 1.4 Billion People

The numbers state the problem clearly. According to the IEA, African data centers consumed roughly 1.4 TWh of electricity in 2024. Europe represented about 15% of global 415 TWh consumption in 2024; the United States roughly 180 TWh. This gap indicates primarily low local electrical activity of data centers, which is an indicator, but not a direct measure, of limited local computing capacity.

To keep pace with global AI deployment, average power demand from African data centers must rise significantly by 2030, according to McKinsey projections. The projected 2030 capacity represents an increase of 3.5 to 5.5 times the current level. Electrical reliability is a major constraint conditioning the realization of this trajectory.

The International Energy Agency estimates that nearly two Africans in five, roughly 600 million people, lack access to electricity. Without sufficient expansion of supply and networks, data center demand can increase electricity tensions and costs. The trade-off is real, not rhetorical. It plays out now, before investments in generation and distribution allow powering both.

The IMF puts the gain African economies can expect from AI under present conditions at 0.2% of GDP. In IMF regional comparisons, the projected gain is roughly 1% in Europe and the Western Hemisphere; the favorable African scenario reaches 2.1% productivity growth over ten years. The direct comparison of current conditions gives roughly a factor of five; a factor close to ten applies only when comparing to the favorable African scenario of 2.1%. This gap depends on infrastructure, connectivity, skills, adoption, and complementary institutions.

A Logic of Capture That Institutions Have Not Yet Countered

Daron Acemoglu, whose 2025 Nobel lecture addressed the link between institutions, technology, and prosperity, poses a question that the African case illustrates acutely: to whom do gains from technological progress belong when local institutions cannot direct it.

Acemoglu emphasizes that automation can generate productivity gains but also displaces tasks, while creating tasks complementary to workers is particularly important for employment and labor’s share. When deployment remains piloted by foreign private actors alone, without public direction of infrastructure orientation, benefits risk concentrating outside affected territories. A significant portion of computing and data services can be supplied from outside, while local spillovers depend on infrastructure, skills, rules, and practices.

The African case is the exact illustration of this mechanism. The Microsoft-G42 project announced in Kenya officially included, beyond the data center, training, an innovation lab, business support, and connectivity investments. They aim to access growing markets and, often, resources, land, energy, data corridors, on favorable terms. AI and advanced computing capacity remains highly concentrated in a few countries and firms, which can create dependencies for African countries.

This reading is not unanimous. Market-liberal economists argue that foreign infrastructure investment, even imperfect, is preferable to no investment. A Microsoft data center in Kenya trains local technicians, creates local suppliers, and generates positive externalities that abstention does not produce. This tension is legitimate: the history of special economic zones in Southeast Asia shows that well-managed foreign investment can indeed trigger upgrading. The difference lies in the word “managed.”

Without an institutional framework capable of structuring counterparts, skills transfer, local content, contribution to public networks, foreign investment in digital infrastructure risks reproducing the enclave structures of extractive industries.

Regional Networks: The Public Trade-off No One Has Yet Made

The solution is not mysterious. It is simply costly and politically difficult.

Interconnection of African electricity networks is an old project. The East African Energy Pool has connected several countries since the 2000s; the West African Power Exchange has functioned partially even longer. But these interconnections cover a fraction of needs. National networks remain fragmented, transmission lines are undersized, and line losses on some networks exceed 20%, meaning a fifth of generated electricity disappears before reaching a consumer.

The African Union inscribed continental electrical infrastructure in its Agenda 2063. According to the World Bank, achieving universal electricity access by 2030 would require roughly $45 billion per year; the International Energy Agency estimates close to $150 billion cumulative for sub-Saharan Africa by 2035. The financing gap for African infrastructure is not new, but AI’s arrival gives it a new dimension: without reliable electricity, Africa cannot simply “skip” the industrial stage to enter the computing economy, as it skipped landlines to move directly to mobile.

Kenya illustrated that this trade-off can be made explicitly. Kenya publicly highlighted the electrical capacity constraint of the Microsoft-G42 project, without establishing a general condition of prior household priority. This is an institutional decision. It asserts that public authority has a role in directing digital infrastructure, that this role cannot be delegated to private investors alone, and that sequence matters: you cannot build an AI economy on a failing electricity grid.

The Infrastructure Africa Is Building

The picture would not be complete without initiatives advancing. Several deserve attention.

Nigeria, the continent’s leading economy, has pursued reforms of the electricity sector since 2024 aimed at improving separation between generation, transmission, and distribution—a condition for attracting private financing in each link of the chain. The stated objective is to reach 30 gigawatts of installed capacity by 2030. The road is long; the country currently operates around 4 gigawatts effective against 12 gigawatts installed, but the regulatory framework is beginning to allow investments that the old monopolistic structure made impossible.

In parallel, submarine cables connecting Africa to the rest of the world are multiplying. Google, Meta, and several consortiums have laid or announced cables that bypass Africa or make landfall there, bringing international bandwidth capacity to unprecedented levels. This connectivity infrastructure does not solve the electricity problem, but it reduces latency and cost of cloud services, allowing access to computing capacity located elsewhere with lower local energy consumption.

Actors like Cloudflare, AWS, and several regional operators are experimenting with distributed computing models that adapt to African network variability: edge network servers, smaller, less power-hungry, capable of running on decentralized renewable sources. Solar is a genuine asset here; African irradiance is among the world’s highest, and several projects couple solar farms with mini-data centers in zones where the national grid does not reach.

These initiatives are real, but their scale does not close the gap. Baker McKenzie, in its June 2026 report on Africa, identifies projects underway in fourteen countries while noting power, transmission, and bankability constraints limiting their sectoral impact. To produce this effect requires sufficient projects in the right places, accompanied by an adapted regulatory framework capable of attracting private investment.

Two Trajectories Before 2030, and What Sets Them Apart

The bifurcation is visible. It plays out in the coming period, before African economies consolidate investment structures and control over their digital infrastructure.

In the first trajectory, African states treat electricity infrastructure as strategic public goods, commit public investment in generation and distribution, and structure regulatory reforms to enable public-private partnerships while maintaining direction control. Negotiations with foreign investors include training counterparts, local content, and public network contributions. In this scenario, AI computing can gradually establish itself locally, data remains on the continent, and potential economic gain increases. Signals to watch: Nigeria’s electricity sector reforms, progression of regional interconnections, and governments’ capacity to impose reciprocity clauses in cloud giant agreements.

In the second trajectory, electrical scarcity persists, computing remains located outside, and African firms access AI via cloud services operated from Europe or the United States. Concentration of AI capacities can increase external dependencies if governments do not develop necessary infrastructure, skills, and governance frameworks.

Indicators of this drift are suspension or abandonment of regional interconnection projects, progression of digital infrastructure agreements without structured counterparts, and insufficient public investment in electricity generation.

What can tip toward the first trajectory is not solely a question of money. Institutions capable of planning and negotiating are an important condition among several necessary foundations for a local computing trajectory, what Acemoglu calls institutional redirection. Kenya showed such a decision was possible, with short-term trade-offs. The question remains open: do these local trade-offs constitute a coherent strategy or remain isolated episodes.

The Trade-off Will Not Wait for the Networks

Africa does not start from zero. It has engineers, growing markets, considerable renewable resources: by 2050, roughly one in four people in the world will be African. These are real assets for the AI economy, provided infrastructure allows converting them.

The electricity problem is solvable. Countries like Morocco and Ethiopia have shown that sustained political will could transform a network in a decade. International financing—World Bank, African Development Bank, climate funds—exists and seeks bankable projects. The solar transition offers a decentralized production pathway that partially bypasses large network limits.

Kenya posed a methodological question: in what order to build, and for whom. Power households first or servers. Train local technicians first or let foreign operators enter without counterpart. These trade-offs happen now and influence trajectories of African digital economies, before the continent has the institutions and networks to fully direct their conditions.

Sources

  1. Semafor, Africa’s Data Center Growth Rests on Power Sector Overhaul (May 6, 2026): https://www.semafor.com/article/05/06/2026/africas-data-center-growth-rests-on-power-sector-overhaul
  2. Daron Acemoglu, Nobel Lecture: Institutions, Technology, and Prosperity, American Economic Review, vol. 115, no. 6, 2025: https://www.aeaweb.org/articles?id=10.1257%2Faer.115.6.1709
  3. International Energy Agency (IEA), Africa Energy Outlook 2025 (annual report, data on electricity access and data center demand)
  4. McKinsey Global Institute, AI in Africa: Unlocking Economic Value (2026 report, electrical demand projections for data centers)
  5. International Monetary Fund, World Economic Outlook 2026 (estimate of AI gains for African economies, 0.2% of GDP)
  6. Baker McKenzie, Africa Data Centre Investment Report (June 2026, status of projects in fourteen countries)
  7. Businessday Nigeria, articles on Nigerian electricity sector reform, 2024-2026
  8. World Economic Forum, reports on digital infrastructure in sub-Saharan Africa