They have returned to the gate. Not with gunboats, Bibles tucked beneath imperial flags, or surveyors measuring land for somebody else's empire. They arrive with cloud credits.
They speak softly about inclusion, innovation, democratisation and "capacity building". They offer universities access to magnificent language models, governments digital transformation, and African start-ups cloud credits. Tools that the continent could never afford to build itself.
Then comes the small print: give us the data. There, in four words, sits the new scramble for Africa. The nineteenth-century bargain was crude. Europe wanted land, minerals, labour and ports. The twenty-first-century bargain is considerably more elegant. The new empire does not necessarily need to own your territory. It merely needs to know your people better than you do.
Data is one of the exponentially growing currencies of the modern world because it does something gold never could. Gold can be spent. Data can be copied, recombined, analysed, trained upon, profiled and repeatedly converted into intelligence. A mineral eventually runs out. A population manufactures fresh behavioural data every hour simply by living. Africa therefore possesses something Silicon Valley cannot manufacture in California: only Africa itself can.
Our languages, diseases, spending habits, crops, weather, classrooms, accents, informal economies, migration patterns, banking behaviour and cultural archives. Our movements, preferences, anxieties and failures. That is not digital exhaust. It is economic ore. And we are in danger of exporting it raw.
We have done this before. Africa exported cocoa and bought chocolate. We exported crude oil and imported refined fuel. We exported copper and imported electronics. We exported diamonds and bought jewellery.
Now we risk exporting human experience and importing intelligence trained upon it. That would be colonialism with a better user experience.
The essential distinction is between access and agency. Access means somebody gives you an account and password. Agency means you decide where your data resides, who may use it, for what purpose and at what price. It means being able to audit what is built from it, retain appropriate intellectual-property rights and participate in the value created downstream.
This is why "free access to AI" should make African policymakers nervous. Free is one of capitalism's most expensive words. When sophisticated digital services are offered cheaply or freely at enormous scale, governments should ask what economic asset exists on the other side of the transaction. Sometimes, it is us.
The asymmetry is extraordinary. The custodian of a valuable dataset may receive research funding, discounted computing or cloud credits. The company extracting patterns from that data may incorporate what it learns into a proprietary system. This is monetised millions of times. One party supplies part of the raw material. The other owns the machine that turns raw material into capital.
Economists have elegant expressions for unequal value capture. Africa has sufficient historical experience to recognise extraction when it sees it.
The sovereignty question is not whether African data should ever cross borders. Of course it should. Science depends upon exchange. Medicine requires collaboration. Commerce needs interoperability. Digital isolation would impoverish rather than protect Africa.
The question is whether value comes back. A fair digital economy requires Africans to retain enforceable rights, bargaining power, technological capability and meaningful participation in the wealth African data helps create. Data sovereignty, then, cannot be reduced to privacy notices and consent boxes. It is increasingly economic sovereignty.
Yet Africa enters the AI age carrying an old structural weakness into a new economy. Fibre may reach our shores, smartphones may fill our pockets, and governments may pass excellent data-protection legislation. But possession of data without sufficient compute, capital, technical capability and commercial ownership can direct the decisive economic power elsewhere.
The old colonial map is being redrawn as architecture. The mine becomes the dataset. The railway becomes the fibre cable. The refinery becomes the model. The chartered company becomes the platform. And once again, the most profitable part of the value chain risks sitting somewhere else.
India offers a useful warning, although not the crude claim that its technology infrastructure produced little wealth. That would be demonstrably false. India's global capability centres now employ roughly 2.36 million people and generate about $98.4 billion annually.
They have created formidable engineering capacity, jobs, expertise and increasingly sophisticated research and product development.
But here lies the lesson Africa should study: infrastructure and technical labour do not necessarily occupy the summit of the digital value chain. Ownership does. Many of these centres ultimately serve multinational enterprises whose parent companies retain control of the global brands, platforms, intellectual property, distribution networks and customer relationships through which value is multiplied. India earns billions, deservedly and increasingly. But the owners of globally scalable digital platforms can command corporate valuations measured in hundreds of billions or even trillions. The comparison is imperfect, but the principle is not: there is money in building and servicing the digital machine; there is considerably greater economic power in owning the machine.
Africa has to ensure that it does not spend the next thirty years proudly counting jobs at the bottom and middle of the value chain while somebody elsewhere owns the top.
Consider Anthropic. Founded in 2021, the AI company reached a valuation of roughly $61 billion by March 2025, while its annualised revenue reached about $3 billion by May that year. Whatever one thinks about feverish AI valuations, capital markets are signalling something Africa would be foolish to ignore.
The extraordinary premium attaches to ownership of the model, the intellectual property, the platform and, ultimately, the relationship with the customer. Africa must hear that scream. We simply cannot become the digital equivalent of a country that owns the mine shaft while somebody abroad owns the mineral rights.
This does not mean technological nationalism. Africa cannot manufacture every frontier model tomorrow. Rejecting useful technology merely because it originated elsewhere would be absurd. Use their models. License them. Partner with their makers. Invite investment. But stop confusing partnership with surrender.
It's time for African governments to treat strategic datasets as economic assets. Public-health information, agricultural records, indigenous-language corpora, geospatial information, educational datasets and public-administration records should not quietly migrate from African control through memoranda decorated with agreeable words such as innovation, inclusion and transformation.
Every major public data partnership should survive five brutally simple questions: Who owns the original data? Who owns the derivative data? Who owns improvements produced from it? Where is the computing performed? And where does the resulting money go?
If those questions cannot be answered clearly, the smiling delegation at the gate should remain outside. Because this is about considerably more than money.
Models trained on African populations may increasingly influence who receives credit, insurance, employment, healthcare or public assistance. If those systems are foreign-owned, opaque and constructed around assumptions Africans cannot adequately interrogate, we may eventually find ourselves governed by automated judgements we neither wrote nor understand.
That is not merely a technology problem. It is constitutional power wearing a hoodie.
The AI divide could simply be laid on top of every previous African divide: electricity, connectivity, education, research, capital and computing. Artificial intelligence may generate extraordinary productivity gains, but countries without the infrastructure, skills and institutional capacity to exploit it will participate principally as consumers.
And consumers pay. Owners collect.
Perhaps Africa needs something less glamorous than another conference entitled "AI for Good". It needs reliable electricity, data centres, high-performance computing and African-language datasets. Serious research funding, regional cloud capacity, procurement rules that reward domestic value creation, and universities producing builders rather than permanent technological tenants are needed as well. Most of all, it needs political leaders capable of saying no.
No to contracts exporting strategic data without reciprocal value. No to "free" platforms that become indispensable before anybody discovers the real price. No to digital philanthropy that improves somebody else's proprietary intelligence while leaving African institutions dependent. And yes, sometimes they can keep their large language models.
Africa does not need to win a childish competition to build the largest general-purpose language model on earth. Africa needs systems that understand African law, agriculture, medicine, logistics, languages, markets and social realities.
The next important African technology company may never attempt to build another ChatGPT. It may instead build the definitive intelligence system for African agriculture, pathology, mining safety, township retail, logistics, education or public administration.
But Africans cannot build tomorrow's industries if we casually surrender today's raw material. For centuries, Africa was told foreigners came because Africa had too little. The terrible joke of history is that they usually came because Africa possessed so much.
Gold. Rubber. Copper. Oil. Diamonds. Human labour. Now add data.
The resource this time is invisible. It is produced every second by ordinary Africans tapping screens, speaking languages, visiting clinics, buying food, borrowing money, attending schools, travelling through cities and merely living their lives.
The harvesters understand its value. They would hardly be at the gate otherwise. The question is whether we understand its value before we sign. The gates are open again. This time, before admiring what the visitors have brought, Africa should ask the question history taught us several centuries too late:
Not merely, "What have you come to give us?" But, "What have you come to take?"