Africa's AI potential is hitting the electricity wall

Artificial intelligence (AI) is often presented through its most spectacular applications, whether it be software capable of producing text, tools to aid in medical diagnosis, or systems automating certain administrative functions. For sub-Saharan African economies, however, the debate is less about access to the most advanced models than about the ability of businesses, governments, and workers to integrate them into their daily operations. Without reliable electricity, affordable connectivity, usable data, and trained employees, the anticipated gains will remain concentrated in a small segment of the economy.

In an analysis published on July 21, 2026, the International Monetary Fund estimates that, at current levels of preparedness, artificial intelligence will add only about 0.2% to sub-Saharan Africa's gross domestic product (GDP) over the next decade. This result, close to stagnation at the regional economic level, contrasts sharply with the most optimistic scenario studied by the institution. If countries improve their digital infrastructure, power supply, skills, cybersecurity, and data governance, the cumulative gain could approach 4% of GDP over ten years, or nearly half a percentage point of additional growth per year.

The difference between these two estimates significantly alters the interpretation of the figure. The IMF does not claim that AI will automatically increase regional wealth by 4%, but indicates that this gain will depend on public policies and investments capable of expanding its use beyond large, already connected companies. The result is based on economic modeling and remains subject to the usual uncertainties surrounding the diffusion of an innovation whose uses, costs, and effects on labor continue to evolve.

The gap between 0.2% and 4% primarily demonstrates that software availability alone is insufficient to generate productivity gains. A company can access a generative AI tool from Dakar, Abidjan, or Nairobi, but it will be unable to use it consistently if the electricity supply is unreliable, the connection is too expensive, or its business data is neither organized nor digitized. At the national level, the widespread adoption of AI therefore depends on investments in energy, telecommunications, education, and public administration, as well as in the IT sector.

The electricity constraint remains the most immediate. The Sustainable Development Goal energy monitoring report, published on June 24, 2026, estimates that more than 560 million people still live without electricity in sub-Saharan Africa. The continent thus accounts for the majority of the world's population without access, while AI models, data centers, telecommunications networks, and business equipment require a continuous power supply.

Even when a business is connected to the grid, the quality of the supply is just as important as the access itself. Outages force production units, businesses, and digital operators to invest in backup generators, batteries, or solar power systems, increasing their costs. In such an environment, adopting new software can improve certain tasks, but it doesn't compensate for the losses caused by machine downtime, grid interruptions, or processing delays.

The connectivity gap creates a second divide. The International Telecommunication Union estimated that by 2024, only 38% of Africa's population would be using the internet, compared to 68% globally and 93% in high-income countries. The continent also exhibits a significant gap between phone ownership and actual network usage, with 66% of the population owning a mobile device while fewer than four in ten were connected.

These figures serve as a reminder that phone coverage is not synonymous with full participation in the digital economy. A basic phone, a limited data plan, or an unreliable connection may allow users to make some payments and exchange messages, but does not necessarily provide access to services requiring more data, computing power, or uninterrupted service. In 2024, the average monthly mobile data consumption in Africa was 3.1 gigabytes per subscription, compared to 13.9 gigabytes globally, according to the ITU.

The spread of AI risks replicating the existing disparities between large corporations and small, informal businesses. Groups with stable access to electricity, structured databases, IT teams, and investment capacity will be able to automate certain functions, better forecast their inventory, or analyze their customer base more quickly. A micro-enterprise operating with paper records, an intermittent internet connection, and limited cash flow will remain marginalized, even if free online tools are available.

However, the productive structure of sub-Saharan Africa relies heavily on small businesses, informal activities, and family farming. The IMF estimates that by 2030, the region will account for approximately half of all new entrants to the global labor market. To have a significant impact on growth, AI will therefore need to improve the productivity of these activities and not remain confined to banks, telecommunications, consulting firms, or subsidiaries of multinational corporations.

In agriculture, tools combining climate data, satellite imagery, and soil information can help optimize planting times, detect certain diseases, and manage irrigation. However, their effectiveness depends on the availability of reliable local data, services accessible in the languages spoken by farmers, and networks that transmit information to rural areas. A relevant agronomic recommendation that is received too late, is incomprehensible to the recipient, or is impossible to implement due to a lack of inputs yields no measurable benefit.

The same reservations apply to public administrations. AI can facilitate the processing of tax documents, the detection of anomalies in public procurement, the analysis of health data, or the management of administrative requests. However, such automation requires reliable records, digitized archives, consistent identifiers, and rigorous protection of personal information. Digitizing incomplete or poorly verified data does not correct its flaws but can, on the contrary, accelerate the production of erroneous decisions.

Training is another prerequisite for widespread adoption. African countries need more than just engineers capable of designing advanced models, although these profiles will remain in demand. Economic gains will also depend on the ability of accountants, agronomists, teachers, doctors, public servants, business owners, and SME managers to integrate these tools into their work. The challenge, therefore, lies as much in continuing education and vocational training as in higher education programs in computer science.

This adaptation of work will need to be managed, as not all effects will be positive. Certain administrative, commercial, or intellectual tasks may be automated, which will alter recruitment needs and income distribution. The most highly trained employees are likely to capture a larger share of the gains, while workers performing repetitive tasks could face increased pressure on their jobs or wages. Therefore, an increase in GDP does not guarantee either uniform income growth or an automatic improvement in working conditions.

States will also have to decide between using imported solutions and developing local capacity. Most of the highest-performing models are designed and hosted outside Africa, meaning that some of the value, data, and computing costs can be captured by foreign companies. This dependence does not negate the benefits of using these services, but it limits local returns if African economies remain solely consumers of software developed elsewhere.

Senegal possesses assets that could support adoption, including a developed telecommunications sector, a digital services ecosystem, and training institutions already working on data-related professions. However, the gap between Dakar and other parts of the country remains significant, as the applications capable of driving national growth must reach agriculture, local government, small businesses, and basic services. The performance of a few digital companies will not be enough to sustainably boost the productivity of the entire economy.

The necessary investments extend beyond public spending alone. Telecom operators, energy producers, banks, universities, businesses, and investment funds will need to contribute to financing networks, data centers, training programs, and solutions tailored to local needs. States will nevertheless retain responsibility for defining rules relating to data protection, competition, cybersecurity, and the use of AI in public services.

The estimate published by the IMF on July 21, 2026, should therefore be interpreted as an indicator of the conditional potential of artificial intelligence, and not as a trajectory already underway. At the current level of preparedness, the estimated cumulative gain of 0.2% would remain very low. The scenario close to 4%, on the other hand, assumes a rapid expansion of access to electricity, the internet, and skills, accompanied by institutions capable of managing the new risks.

The African debate on AI therefore largely hinges on the budgets allocated to electricity grids, fiber optics, education, research, and the digitization of public administration. While the tools themselves can evolve in a matter of months, the infrastructure and skills required to use them take years to build. It is this time lag that will determine the continent's ability to transform imported innovations into widely distributed productivity gains.