African governments must pursue a continental computational sovereignty agenda, writes Fulufhelo Nelwamondo
Since the 15th century, when European colonial expansion began to reorder the world, powerful nations extracted raw materials from weaker societies, shipped them abroad for processing, manufactured high-value products elsewhere, and sold them back to the very countries from which those resources originated. Gold left Africa as ore and returned as jewellery. Oil left producing regions as crude and returned as petrochemicals. Rubber, cotton, copper, diamonds, and cobalt followed the same logic: extraction here, intelligence elsewhere, profit somewhere else.
The AI age is reproducing that model, except this time the resource is not gold, oil or diamonds.
It is data. Africa is once again in danger of occupying the bottom of the value chain. The world has become mesmerised by the promise of artificial intelligence: productivity gains, automation, smarter government, medical breakthroughs, autonomous systems, scientific acceleration, and economic transformation. But beneath the glossy language of innovation lies a far more uncomfortable reality. AI is not merely creating technological inequality. It is reorganising global power itself.
The nations and corporations that control algorithms, compute infrastructure, cloud platforms, chips, and foundational AI models are positioning themselves to become the new imperial powers of the 21st century. Those that merely consume these systems risk becoming digital colonies, permanently dependent on foreign intelligence systems they neither own, shape, inspect nor fully understand.
This is not a metaphor. It is an economic architecture already taking shape. Professor Tshilidzi Marwala’s recent United Nations University article, ‘The Algorithm Divide: Why Developing Nations Risk Being Left Behind Forever’, correctly warns that development is increasingly being determined by access to the “algorithm economy”, with national wealth measured not only in factories and minerals, but in data and compute. But the danger is even sharper than being left behind.
The greater danger, however, is dependency. Today, billions of Africans generate enormous volumes of behavioural data every second: search histories, mobility patterns with exact GPS coordinates, financial transactions, social interactions, consumer preferences, voice recordings, biometric traces, health records, education data, and political signals. Yet the overwhelming majority of platforms collecting, storing, processing, and monetising this data are foreign-owned.
African societies increasingly function as suppliers of raw digital material, while the intelligence derived from that material, the algorithms, predictive systems, models, patents and profits, accumulates elsewhere. Colonialism extracted minerals. The AI age extracts cognition. That is the uncomfortable truth many policymakers are still unwilling to confront.
Historically, sovereignty was measured through control over territory, borders, armies, currencies, and natural resources. Increasingly, national power is being determined by something less visible but more decisive: data infrastructure, computational capacity, AI talent, semiconductor access, cyber capability, cloud sovereignty, and algorithmic influence.
Countries without these capabilities will not simply “fall behind”. They risk becoming structurally dependent. Governments will rely on foreign AI systems to manage public services. Banks will depend on offshore algorithms for credit scoring and fraud detection. Healthcare systems will use imported diagnostic intelligence trained on populations unlike their own. Education platforms will shape learning through recommendation engines designed elsewhere. Defence institutions will acquire autonomous systems whose deepest logic they cannot inspect. Entire economies may eventually operate on computational infrastructure owned by companies headquartered thousands of kilometres away. At that point, dependency stops being technological. It becomes civilizational.
The first layer of this dependency is cloud dependence. Synergy Research Group reported that Amazon, Microsoft, and Google together accounted for about 63% of global enterprise spending on cloud infrastructure services in Q3 2025. In other words, much of the world’s digital economy increasingly runs on infrastructure controlled by three American companies. This matters because cloud is no longer merely where companies store documents. It is where intelligence is trained, deployed, monetised, and governed. Africa’s position in this emerging order is worrying. The continent’s data centre capacity remains a small fraction of global capacity. Recent reporting on Africa’s data centre landscape estimated Africa’s active data centre capacity at about 360MW, compared with global active capacity of about 55GW. This is the infrastructure version of colonial dependency: our data grows here, but the intelligence is processed elsewhere.
The second layer is chip dependence. AI runs on hardware: graphics processing units, specialised accelerators, advanced semiconductors, networking equipment, memory systems, and energy-hungry data centres. The countries that dominate chips will dominate the tempo of AI development. Reuters has reported on massive AI infrastructure deals involving companies such as Nvidia, AMD, and OpenAI, including arrangements worth tens of billions of dollars to supply chips and build AI data centre capacity. The AI world is a capital-intensive industrial race, and that has profound implications for Africa.
If the continent has no serious semiconductor strategy, no sovereign compute agenda, and no continental approach to AI infrastructure, then it will remain a buyer in a market where others set the prices, standards, access conditions, and geopolitical rules.
The third layer is model dependence. Foundation models are becoming the invisible infrastructure of decision-making. They write text, generate code, summarise legal documents, screen job applications, support medical interpretation, power chatbots, assist engineers, analyse images, and increasingly mediate human interaction with knowledge. But most foundation models are trained outside Africa, in languages, cultures, legal systems, and institutional realities different from ours. They are optimised for dominant markets. They reflect dominant data. They encode dominant priorities.
A society that relies on imported intelligence systems must ask difficult questions. Whose histories are overrepresented? Whose languages are marginalised? Whose legal assumptions are embedded? Whose risk models determine access to credit? Whose cultural references define “normal”? Whose values shape automated advice? Whose errors become our administrative reality? That is how dependency becomes epistemic.
The fourth layer is data extraction. United States and China accounted for about 90% of the market capitalisation of the world’s largest digital platforms, and that the top global platforms were investing across the global data value chain to strengthen their competitive advantage. This is the digital equivalent of controlling mines, railways, ports, refineries, and markets at the same time. The platform sees the user, captures the data and processes the data, trains the model, and then sells the service. The platform captures the margin while the user becomes both consumer and raw material.
Africa must take this seriously because the continent is one of the world’s fastest-growing digital frontiers. GSMA’s work on Sub-Saharan Africa shows expanding mobile connectivity and rising mobile data traffic, with mobile data traffic per connection expected to grow sharply by 2030. This growth also means the continent is generating more data than ever before. The question is: who captures the value?
The fifth layer is research dependence. AI research itself is becoming less democratic. The Stanford AI Index’s 2025 report highlights the scale and acceleration of AI progress, investment, and adoption. Academic work on the “compute divide” has also warned that modern AI research increasingly favours large firms and elite universities with access to specialised computing resources, crowding out less-resourced institutions. This matters deeply for African universities. If serious AI research requires computational resources that only a few global corporations and elite institutions can afford, then African researchers may be reduced to spectators in the production of frontier knowledge.
The sixth layer is regulatory dependence. Many African governments are rushing to draft AI policies filled with fashionable language: ethics, inclusion, safety, transparency, fairness, and responsibility. These principles are important. But they can easily become decorative if they are not linked to infrastructure, industrial policy, and state capability. A country cannot meaningfully govern technologies it does not understand, cannot inspect, and does not control. That is not sovereignty. It is outsourced intelligence.
Where is the sovereign compute infrastructure? Where is the national AI industrial strategy? Where is the large-scale public data infrastructure? Where is the semiconductor roadmap? Where is the African language AI programme? Where is the continental cloud strategy? Where is the energy plan for AI infrastructure? Where is the mathematics, statistics, and computational talent pipeline?
A country cannot participate seriously in the AI era while treating computational infrastructure as a luxury instead of strategic national infrastructure. Artificial intelligence will influence economic competitiveness, labour markets, military superiority, intelligence systems, scientific discovery, media ecosystems, political persuasion, public administration, and cultural identity. The nations that dominate AI will increasingly shape how humanity thinks, communicates, trades and governs itself.
That is why the language of “digital colonies” is not alarmist. It may in fact be too polite. Traditional colonialism controlled land and labour. Algorithmic dependency may ultimately control cognition itself. Recommendation systems already influence what populations read, what they buy, what they believe, how they vote, and how they perceive reality. When those systems are externally owned, externally trained, and externally governed, dependence moves beyond economics into the realm of societal autonomy.
This is the central warning: the AI age will not only divide countries between rich and poor. It may divide them between those who own intelligence and those whose lives are governed by intelligence owned elsewhere. Africa must therefore urgently decide what role it wishes to occupy in the global AI value chain.
This does not mean Africa must replicate Silicon Valley or compete directly with OpenAI, Google DeepMind, Anthropic or China’s frontier model ecosystem. That would be unrealistic in the short term. But it does mean Africa must identify strategic areas where sovereign capability is essential.
In my view, Africa should prioritise public-interest AI, multilingual African language models, agricultural intelligence, mining optimisation, healthcare diagnostics, education systems, public administration, climate resilience, scientific discovery, and infrastructure planning. The continent does not need to dominate every AI frontier. It must dominate relevance.
In the 19th century, nations built railways. In the 20th century, they built electricity grids, ports, highways, and telecommunications systems. In the 21st century, serious nations will build computational infrastructure. Compute is not an ICT accessory. It is the new strategic infrastructure. African governments should therefore pursue a continental computational sovereignty agenda, which must include sovereign cloud capacity, regional data centres, national research compute facilities, public-sector data trusts, open African language datasets, AI regulatory sandboxes, procurement policies that support local AI firms, and massive investment in mathematics, engineering, statistics, and computer science.
Fulufhelo Nelwamondo holds a PhD in Electrical Engineering, is a member of the South African Institute of Electrical Engineers, a senior member of the Association of Computing Machinery, as well as the Institute of Electrical and Electronics Engineers.