Every week, a new AI startup emerges somewhere in East Africa. They’re building legal assistants, healthcare platforms, fintech copilots, agricultural tools, customer service agents and education platforms. By every measure, the region’s AI ecosystem is growing. But beneath the excitement lies a question we rarely ask:
Where does the intelligence actually come from?
The answer reveals one of the biggest strategic questions facing Africa’s digital future. Most AI companies in East Africa do not own the intelligence powering their products. They build applications on top of large language models developed by companies such as OpenAI, Anthropic, Google, Meta, Mistral and others. Every time a user asks a question, the application sends that request to a model running in hyper-scale data centers, typically in the United States, Europe or South Africa. Within seconds, the model processes the request and sends back a response.
To the user, it feels like the AI lives inside the application. In reality, the application is renting intelligence from somewhere else. There is nothing inherently wrong with this. It is the fastest and most cost-effective way to build AI products. Training a frontier model requires billions of dollars, enormous GPU clusters, vast amounts of electricity and years of research. Very few companies in the world can do it.
The real issue is that we have mistaken building AI applications for building an AI industry. These are not the same thing. An AI application economy creates products. An AI industry owns the infrastructure that powers those products. It owns the compute, the cloud platforms, the GPU clusters, the data centers, the networking infrastructure and, increasingly, the models themselves.
Today, most of East Africa’s AI ecosystem sits at the application layer. Our engineers build the software. Our businesses create the demand. Our citizens generate the data. But the reasoning happens elsewhere. That has important economic consequences.
Every AI request made by an African business is processed on infrastructure owned by another company. Every response generates demand for someone else’s cloud, someone else’s GPUs and someone else’s data center. While African startups capture value through innovation and customer relationships, much of the underlying infrastructure value accrues to the companies that own the AI stack.
This is why infrastructure matters.
South Africa understood this early. Companies like Teraco did not simply build data centres. They built digital infrastructure that attracted AWS, Microsoft Azure, Google Cloud, Oracle and hundreds of network providers. Once the infrastructure existed, investment, cloud services and AI workloads followed.
East Africa has talented engineers, ambitious founders and a growing startup ecosystem. What it lacks is infrastructure at that scale. If the region wants to compete in the AI era, the conversation must evolve beyond AI courses, hackathons and startup competitions.
We should also be asking: Who will build East Africa’s AI-ready data centers? Who will finance regional GPU infrastructure? Who will host African AI models? Who will provide sovereign AI cloud services? These are no longer technology questions alone., They are questions of economic strategy, industrial policy and long-term competitiveness. The countries that produce the next generation of AI applications will create valuable companies, The countries that own the infrastructure beneath those applications will shape the future of the AI economy. For East Africa, the next frontier is not simply adopting artificial intelligence. It is building the infrastructure that artificial intelligence depends on.
James Kaliisa is Co-Founder & CTO of Nexus Inc., a Rwandan deep-tech company building sovereign artificial intelligence infrastructure and digital systems for Africa.
