Africa’s artificial intelligence (AI) debate often begins with applications. How can AI improve agriculture? Can it help doctors diagnose diseases, teachers prepare lessons or governments deliver services more efficiently?
These are important questions, although they begin too far down the chain.
Before an AI application can advise a farmer, analyse a medical record or process a government document, it needs computing power. That computing power sits in data centres that require reliable electricity, internet connectivity, cooling systems, advanced equipment and skilled workers. The application also depends on data, often stored and processed through platforms controlled by companies outside Africa.
The first question, therefore, is not simply what Africa can do with AI. It is who will own, power and govern the systems through which that AI is delivered.
The issue has become urgent because adoption is moving quickly. The 2026 Stanford AI Index reports that generative AI reached approximately 53% level adoption within 3 years of its mass market introduction, faster than the early adoption of personal computers and the internet.
The economic estimates are equally striking. McKinsey calculates that generative AI could create between US$2.6 trillion and US$4.4 trillion in value annually across 63 use cases. For Africa, the report estimates that large-scale deployment could unlock between US$61 billion and US$103 billion across varied economic sectors.
African governments are right to take this opportunity seriously. The question is not whether the continent will adopt AI. Adoption is already taking place, and staying outside the technology would carry its own economic costs.
What remains unsettled is how much of the value Africa will retain.
The adoption gap is already visible
Microsoft’s estimates for the second half of 2025 show how unevenly generative AI is spreading. Among the working-age population, estimated adoption was 64% in UAE and 60.9% in Singapore. It stood at 21.1% in South Africa, 9.3% in both Ghana and Nigeria, 8.1% in Kenya and 6.8% in Ethiopia.
The figures do not measure the full strength of a country’s AI industry. They tell us little about whether local companies are creating models, whether universities have access to advanced computing or whether governments can regulate the systems being used. They reveal a familiar pattern: countries with higher incomes, reliable electricity, widespread internet access and stronger digital capabilities are moving faster.
AI is not arriving on a level playing field. It is being added to economies that already differ sharply in infrastructure, skills, investment capacity and institutional strength.
The major powers understand that control of AI requires more than high adoption. The United States is pursuing leadership through private investment, infrastructure, innovation and national security. China is expanding its “AI Plus” programme across industry and government while strengthening domestic technological capability. The EU is combining investment with risk-based regulation, while India is expanding access to computing, datasets and indigenous AI development.
Africa has also set out its ambitions. The African Union’s Continental Artificial Intelligence Strategy, adopted in 2024, calls for local capability, stronger infrastructure, responsible use and greater African control over strategically important data. It supports national and regional data pools, clearer rules for cross-border data transfers and stronger data governance frameworks.
The strategy points in the right direction. The harder task is deciding what countries should build nationally, what should be developed regionally, and what can safely remain in the hands of foreign providers.
Africa can use AI without owning much of it
AI is often treated as software that an organisation can buy and add to its operations. The industry is far larger. It includes semiconductors, servers, cloud platforms, data centres, electricity, internet, technical expertise, training data and the applications that users eventually see.
Investment in these resources is concentrated in a small number of countries. Private AI investment in the US reached an estimated US$285.9 billion in 2025. The US also hosts more than 5,400 data centres and remains the leading centre for frontier-model development and advanced AI infrastructure.
Africa’s data centre industry is growing, particularly in South Africa, Egypt, Kenya and Nigeria. The continent is not starting from zero, although the Africa Data Centres Association estimates that it accounts for only 0.6 per cent of global capacity.
This gap does not prevent Africans from using AI. A bank in Lagos can deploy a foreign model to detect fraud, a hospital in Nairobi can subscribe to a cloud-based diagnostic tool, and a government department in Accra can automate the processing of documents. However, the question that arises is where the money goes.
If the cloud platform, model, computing infrastructure and intellectual property are all owned abroad, African organisations pay for access while the most valuable parts of the industry remain elsewhere. Local businesses can still benefit from productivity gains, although the continent captures less of the infrastructure revenue, technical expertise and intellectual property generated by the technology.
Foreign dependence is not a failure. No country controls every part of the AI value chain, and attempting to reproduce the entire American or Chinese system would be unrealistic. The concern arises when governments and businesses depend on services they do not fully understand, cannot audit and have little power to replace.
AI policy is also electricity policy
The infrastructure question is an electricity question.
The International Energy Agency estimates that data centres consumed approximately 415 terawatt-hours of electricity in 2024, equal to about 1.5 per cent of global consumption. Its central projection puts demand at roughly 945 terawatt-hours by 2030, with electricity use by the servers that support AI growing faster than conventional computing.
Sub-Saharan Africa enters this era with a large electricity deficit. Access stood at approximately 55.1% in 2024 and in 2023, 565 million people remained without electricity. Sub-Saharan Africa accounts for 85% of the world’s population without electricity.
Meanwhile, governments are being asked to attract electricity-intensive data centres when households, schools, hospitals and businesses continue to face unreliable supply and high costs.
Data centres can support investment, improve local cloud services and expand computing capacity. Their contribution will need to be judged against their demands on electricity, water and public finances.
Before granting a tax concession or preferential electricity tariff, a government should establish how much power it will consume, where the supply will come from and whether the operator will add new generation capacity. It should also determine how many permanent jobs will be created, how much tax will be paid and whether local companies, researchers and public institutions will receive affordable access to the infrastructure.
Where a data centre company needs dedicated power, the project should expand the electricity system rather than take supply from existing users. Without these conditions, a government could subsidise a global technology company while local businesses continue to pay high tariffs for an unreliable service.
Every country does not need its own large data centre
The desire to be seen as an AI hub can also lead governments to pursue projects that their economies cannot support.
A large data centre requires more than land and a government announcement. It needs consistent electricity, reliable internet, physical security, maintenance and enough demand to justify the investment. Countries without these conditions risk building costly and underused infrastructure.
Regional computing hubs offer a more practical option in Africa. A smaller number of well connected facilities could serve several markets, spread costs and make better use of expensive computing equipment. Smaller countries can purchase capacity without financing an entire national facility.
This approach introduces difficult questions of its own. Countries would need to agree on ownership, access, pricing, cybersecurity and legal jurisdiction. They would need procedures for service interruptions, political disagreements and changes in national law.
Those complications are real, although they are not an argument against regional cooperation. They are an argument for negotiating the rules before sensitive government and commercial data are placed in shared systems.
Where is the data, and whose laws apply?
The discussion about infrastructure cannot be separated from digital sovereignty.
African governments and companies will continue to use foreign cloud platforms and AI systems. The central problem is not that these services come from abroad; it is that many organisations do not know enough about how their information is stored, processed and reused.
A ministry can purchase an AI system to analyse health records, tax returns, migration files or social-protection applications without fully understanding where the data is located. Officials might not know which subcontractors can access it, whether it can be used to improve the vendor’s model or whether a foreign government can obtain it under its domestic laws.
Data protection laws can partly address the problem but not all of it. Governments need to classify information according to sensitivity and establish where different categories can be stored and processed.
Routine administrative data can often be handled by providers under strong contractual and security safeguards. Sensitive information involving national security, health, taxation, or citizens’ identities deserves stricter rules, including approved national or regional storage where necessary.
Public institutions should conduct risk assessments before buying AI systems. Contracts should include cybersecurity standards, incident reporting obligations, audit rights and clear restrictions on the use of government data for commercial model training.
Africa does not need to build the next ChatGPT
An African AI strategy should not attempt to reproduce the entire infrastructure of the United States or China. Most countries do not need to train frontier models, and few can justify the cost.
The better approach is to identify the parts of the value chain that can be built competitively at national or regional level. These include specialised data centres, local datasets, cybersecurity services, model testing and applications developed for sectors in which African businesses and institutions have relevant knowledge.
Smaller and specialised models also offer a practical route. A system built to support agricultural officers, interpret a national tax law or assist a public-health programme does not need to perform every task. It needs to perform its specific function accurately, securely and at an affordable cost.
The objective is not complete technological independence. It is enough capacity to retain economic value, protect sensitive information and negotiate with international providers from a stronger position.
Africa will adopt AI. The outcome now depends on whether governments treat it only as a market for applications or as an industry built on electricity, computing, data and policy choices.
Without stronger infrastructure and governance, African countries can become active users of artificial intelligence while remaining minor producers of its value.
Sources and further reading
Stanford Institute for Human-Centered Artificial Intelligence, The 2026 AI Index Report.
Microsoft AI Economy Institute, Global AI Adoption in 2025: A Widening Digital Divide, January 2026.
McKinsey Global Institute, The Economic Potential of Generative AI: The Next Productivity Frontier, June 2023.
Kuyoro, Mayowa, Umar Bagus, Anass Bensrhir and Ziyaad Bobat. “Leading, Not Lagging: Africa’s Gen AI Opportunity.” McKinsey & Company, 12 May 2025.
International Energy Agency, Energy and AI, 2025.
World Bank, Access to Electricity (% of Population): Sub-Saharan Africa, World Development Indicators.
IEA, IRENA, UNSD, World Bank and WHO, Tracking SDG7: The Energy Progress Report 2025.
African Union, Continental Artificial Intelligence Strategy, July 2024.
African Actors of Data Center Association, Data Centres in Africa 2026: The Economic Report.
Leave a Reply