This week, the Radar stops at three points across the continent. They do not form a picture of ‘Africa’ as a single technological space. They are three distinct scenes which, placed side by side, make visible something that tends to disappear when we talk about artificial intelligence as though it lived only on a screen.
In Cape Town, the debate is about the land, electricity and water demanded by infrastructure. In Egypt, a memorandum with Intel proposes extending AI literacy to one million people every year. In Ghana, the debate over technological sovereignty is translated into a concrete investment in computing capacity.
Infrastructure, learning and sovereignty. Three ways of observing how AI acquires territory.
Cape Town · Infrastructure before the cloud
On 14 July, Cape Town’s Municipal Planning Tribunal approved, by four votes to one, a land-use application that allows progress towards one or two large data centres at King Air Industria, near the airport. The application involves rezoning, subdivision and consolidation of land. Wally Johnstone, the only member of the tribunal to vote against it, argued that he could not support an application of this scale without knowing its water and electricity requirements, and warned of the pressure the infrastructure could place on services already under strain. [1]
In mid-September, Housing Assembly and the British technology-justice organisation Foxglove, represented by the Legal Resources Centre, lodged an appeal against the decision. They argue that the available documentation does not allow an adequate assessment of electricity and water consumption, emissions, noise, backup generators or the cumulative consequences for nearby communities. [2]
The suspension is not a new political decision by the City. The approval remains suspended while the appeal is being processed and cannot be implemented until a final decision is reached. The municipality is also working on specific guidelines for future applications involving large data centres. [3]
Water has become the most visible figure in the dispute, and also the one that requires the greatest caution. Before the tribunal, the Legal Resources Centre calculated that facilities with an electricity demand close to 174 MW could use around 4.4 billion litres a year if consumption associated with conventional evaporative cooling systems were applied. The application documents did not specify how much water the facilities would require, nor did they provide sufficient detail about their electricity demand and supply. [1]
The majority of the tribunal took a different view. Its other four members, including chair Sydney Holden, considered that water and electricity demand, along with the other outstanding issues, could be addressed later during preparation of the Site Development Plan. [1]
That disagreement shifts the dispute beyond a single number. What matters is how much a facility will ultimately consume, but also at what point in an urban approval process its cooling system, energy demand and material effects should have to be known. When those data are deferred to later stages, the very process of producing — or not producing — information becomes part of the infrastructure.
The 4.4 billion litres are not an Equinix forecast. The company has confirmed that the facilities would use dry-air cooling, a technology that virtually eliminates direct water use for cooling server rooms. [4]
The alternative does not remove the material dimension. Dry cooling uses more electricity and can shift part of the water footprint towards the source where that electricity is generated. The dispute therefore moves away from a single figure and towards a chain of interdependent resources. [4]
There is a necessary counterpoint. Obinna Isiadinso, Global Lead for Data Centers and Cloud at the International Finance Corporation, argues that protests should not halt the expansion of this infrastructure across the continent. Africa has around 409 MW of operational capacity, less than 1% of the global total despite being home to roughly one fifth of the world’s population. [5]
The need for infrastructure and the demand to know its costs do not cancel each other out.
Antina von Schnitzler, Associate Professor of International Affairs and Anthropology at The New School, showed in Democracy’s Infrastructure (2016) how apparently ordinary technologies — in her case, prepaid water meters in South Africa — can become sites where citizenship, rights and relations with the state are negotiated. Brian Larkin, Professor of Anthropology at Barnard College, Columbia University, argued in 2013 that infrastructures should not be studied as neutral backdrops to social life, but as material forms that organise circulation, power and possibilities for action. [6][7]
A data centre can be read in the same way. It is not the cloud. It is a concrete way of distributing resources, access and costs.
Egypt · Teaching AI literacy to one million people
On 16 September, Egypt’s Ministry of Communications and Information Technology and Intel signed a memorandum of understanding with an ambition unusual in scale. They aim to train one million people a year for three years in artificial-intelligence concepts, the safe and responsible use of technology, and digital trust. [8]
The agreement also provides for 1,500 certified trainers, five hundred each year. The programmes are intended to reach children and young people, technical and university students, people in work or seeking employment, and public-sector officials. [8][9]
The announcement should not be confused with a programme already deployed to three million people. The parties still have to develop a phased implementation plan. The first initiative will be AI for Citizens, an Intel programme. The collaboration also includes workshops on the company’s infrastructure and computing technologies, an annual hackathon and the preparation of projects for submission to the Intel AI Global Impact Festival. [8][9]
That detail changes how the agreement can be read. Intel does not appear merely as an external partner. Part of the pedagogical architecture comes from its own programmes and technologies. The economic terms of the collaboration are not detailed in the public sources consulted.
Technological literacy is never simply the transmission of technical knowledge. It also selects practices, vocabularies and criteria of use. People learn where to ask, what to delegate, which results to regard as reliable and what behaviour is presented as responsible.
The announced scale allows the discussion to move beyond the familiar question of job displacement. Egypt has signed a memorandum that sets out an ambition to turn a certain level of AI knowledge into a widely distributed competence. What remains unclear is how deep that training will be, how it will materialise and whether learning to use these systems will also include learning how to question them.
There is a continuity here with Cape Town. Millions of people may acquire digital skills, but those skills depend on devices, connectivity, data centres and computing capacity.
Literacy has infrastructure too.
Ghana · Having a voice requires computing capacity
In September, Riyadh hosted UNESCO’s fourth Global Forum on the Ethics of Artificial Intelligence. The meeting addressed governance, institutional capacity, infrastructure, cultural and linguistic diversity, and participation by countries occupying very different positions within the global technological ecosystem. [10]
There, Ghana’s Minister for Communication, Digital Technology and Innovation, Samuel Nartey George, called for a greater presence of African data, human capacity, infrastructure and cultural perspectives in AI systems and in the international rules that govern them. [11]
George linked that claim to concrete infrastructure. Ghana is committing US$250 million over two years to develop AI computing capacity. The stated intention is that the infrastructure should not serve Ghana alone. The minister explicitly referred to access for universities, research, start-ups and neighbouring countries, and argued for regional centres that would avoid every state having to build extremely expensive facilities separately. [11]
In another intervention during the same forum, George said that Cabinet had approved the US$250 million for a national AI and computing centre intended to provide sovereign and affordable access for researchers, universities, start-ups and other technology initiatives. [12]
The figure did not first appear in Riyadh. When the National Artificial Intelligence Strategy was launched on 24 April, the Government had already announced a US$250 million investment in a computing centre and linked that infrastructure to research, innovation and local access to AI tools. [13]
Infrastructure does not appear separately from training. In the same intervention, George pointed to the One Million Coders programme, through which Ghana aims to train one million people in digital skills including programming, artificial intelligence, data security and cybersecurity. The programme is also being extended to universities and secondary schools. The boundary between the Egyptian and Ghanaian cases therefore becomes less distinct: literacy and computing capacity are beginning to be built at the same time. [11]
The sovereignty described by the minister does not mean self-sufficiency either. George explicitly cited Nvidia and Huawei to argue that an African country acting alone has little bargaining power with major technology suppliers, whereas a coordinated position could alter that relationship. ‘Sovereign’ infrastructure, then, will still be built within a network of corporate and geopolitical dependencies that also deserves scrutiny. [11]
The minister also used a revealing example. He suggested that an AI system trained without sufficient local context might interpret a Western woman in a bikini one way while classifying an African woman taking part in a traditional ritual as nudity. It was a hypothetical example intended to show how automated classification systems incorporate the data, conventions and values from which they are built. George then linked that reflection to the governance of social media and the definition of global standards. [11]
The example deserves a second reading. Speaking of ‘African values’ in the singular can replace one bias with another simplification. The societies of the continent do not constitute a single moral or cultural unit, and deciding which practices should represent those values is itself part of a political dispute.
That qualification does not weaken the criticism of data bias. It makes it more demanding.
Abeba Birhane, Assistant Professor of AI and founder of the AI Accountability Lab at Trinity College Dublin, described in 2020 as algorithmic colonisation the importation of technologies built around external priorities and then presented as universal solutions. Her argument calls for systems rooted in local needs and interests. [14]
Sabelo Mhlambi approached the issue from another angle in From Rationality to Relationality (2020), written while he was a Technology and Human Rights Fellow at the Carr Center for Human Rights Policy and a fellow at Harvard’s Berkman Klein Center. Drawing on ubuntu, he proposed a relational ethics of artificial intelligence that shifts attention away from individual autonomy towards dignity, interdependence and the effects of automated decisions on social relations. [15]
These are different perspectives. That is precisely why they are more useful than the idea of a single African cultural response to AI.
Technological sovereignty therefore appears less as a declaration than as a chain of capacities: data, languages, knowledge, electricity, data centres, trained people and access to computing.
Three places, one tension
Cape Town and Ghana reveal two ends of the same chain. In one, organised communities are demanding to know the material costs before particular land-use rights become consolidated. In the other, a government argues that the lack of infrastructure limits the possibility of participating in the AI economy from a less dependent position.
The positions are not incompatible. The roughly 409 MW already in operation point to a genuine shortage. At the same time, the South African dispute is a reminder that correcting that shortage redistributes territory, electricity, water, investment and power.
Egypt and Ghana’s One Million Coders programme add another layer. Building technological capacity is not only a matter of installing servers. It also means deciding who learns to use these systems, with which tools, under which curricula and with what possibilities for questioning them.
That is why the questions already being asked in Cape Town can also accompany the expansion planned in Ghana. How much energy will it require? Which cooling systems will it use? What consumption data will be public? Who will be able to access the computing capacity? Which suppliers will it depend on? Which benefits will remain in the territory, and who will bear the costs?
There is no single continental answer to these questions. There are different negotiations that are beginning to intersect.
Across several parts of the continent, artificial intelligence is ceasing to be discussed only as software.
It is beginning to be negotiated as social infrastructure.
References
- Kretzmann, S. (2026, 14 July). Two giant data centres secure initial green light from Cape Town municipal tribunal. Daily Maverick / GroundUp.
- Foxglove. (2026, September). Civil society appeal against Cape Town hyperscale data centre over critical planning information gaps.
- Van Rooyen, F. (2026, September). Cape Town to write new rules for big data centre applications. TechCentral.
- MyBroadband. (2026, 8 August). False information about new Cape Town data centre’s 4.4-billion litre water usage.
- Kupemba, D. N. (2026, 17 September). South Africa joins the global resistance against American data centers. Rest of World.
- von Schnitzler, A. (2016). Democracy’s Infrastructure: Techno-Politics and Protest after Apartheid. Princeton University Press.
- Larkin, B. (2013). “The Politics and Poetics of Infrastructure”. Annual Review of Anthropology, 42, 327–343.
- State Information Service of Egypt. (2026, 16 September). Egypt, Intel sign MoU to train one million citizens annually in AI.
- Ahram Online. (2026, 16 September). Egypt, Intel to train 1 million citizens annually in AI under three-year deal.
- UNESCO. (2026). 4th Global Forum on the Ethics of Artificial Intelligence. Riyadh.
- 3News. (2026, 15 September). Ghana pushes for African approach to AI ethics and infrastructure.
- 3News. (2026, 15 September). We are closing all gaps in our AI and digital infrastructure – Sam George.
- Ghana News Agency. (2026, 25 April). Ghana to invest $250m in AI Centre, launch national strategy.
- Birhane, A. (2020). “Algorithmic Colonization of Africa”. SCRIPTed, 17(2), 389–409.
- Mhlambi, S. (2020). From Rationality to Relationality: Ubuntu as an Ethical & Human Rights Framework for Artificial Intelligence Governance. Carr Center Discussion Paper 2020-009, Harvard Kennedy School.