The sovereignty question: can Africa build AI without big tech?

Read the national AI strategies of Kenya, Nigeria, Egypt, South Africa, Ghana or Rwanda and a common word appears: sovereignty. Governments want control over their data, their digital infrastructure and the AI systems that increasingly shape their economies. Some policies explicitly flag reliance on foreign technology companies as a risk.

Then read the announcements: partnerships with Google, Microsoft, Nvidia, Meta, Amazon and Chinese firms for data centres, skills programmes and cloud services. A Rest of World investigation this year found that many African AI plans still depend heavily on these companies.

Is that a contradiction? Not necessarily. But it is a tension worth understanding.

What sovereignty could mean

"AI sovereignty" is used loosely. In practice it can mean several different things:

• Data sovereignty: data about citizens and businesses stored and processed under local law
• Infrastructure: data centres and compute located in the country or region
• Capability: local engineers, researchers and companies able to build and adapt AI
• Governance: the ability to set and enforce rules on how AI is used
• Cultural sovereignty: AI that understands local languages, values and contexts

A country can have some of these without others. A data centre run by a foreign company in Nairobi gives infrastructure and some data control, but not necessarily capability.

Why partnerships still make sense

Building frontier AI from scratch requires enormous capital and talent that few countries anywhere have. Partnering with global companies can bring investment, skills and access to tools faster than going alone. The question is not whether to partner, but on what terms:

• Does the partnership transfer skills and create local jobs at senior levels?
• Is data governed by local law, with clear limits on its use?
• Are there local companies in the value chain, not just as resellers?
• What happens if the partner changes strategy, as Microsoft's changes in Nairobi this month show can happen? What the Nairobi cuts mean

The overlooked layer

Sovereignty debates focus on data centres and laws. The layer that touches most citizens is the applications they use every day: the tools that write, translate, design, advise and sell. If those are all built elsewhere, a country can own the data centre and still have little say in how AI shapes its economy.

African-built applications that understand African languages and markets are part of sovereignty too. The year of African-language AI.

Our view

Tari Studio is built in Kenya and uses AI models from different providers, including our own AK1 and AK2, tuned with an AI director that writes African scenes, people and languages into every request. We do not pretend that a small company can build everything itself. We do believe that the parts closest to the customer, understanding the market, the language and the culture, should be built here.

Sovereignty, in the end, is less about owning every layer than about having real choices, and the skills to use them.