Investors want moats, not prompts: the state of African AI funding

Money is flowing to African AI start-ups, but carefully. StartupList Africa tracks 190 AI start-ups on the continent with about $370 million in disclosed funding, led by Egypt, Nigeria and Kenya, and weekly deal roundups such as Techloy's show AI and electric mobility among the most active areas.

But the tone among investors has shifted. In an October opinion piece on TechCabal, Emmanuel Adegboye of Madica argued that investors want AI businesses that cheap code cannot copy, and that capital is concentrating in a handful of funds and start-ups.

The "wrapper" problem

In the first wave of generative AI, many start-ups were built as thin layers on top of a large model from someone else: a nice interface, a few prompts, a subscription. Some made money quickly. Many found that the model provider, or a competitor with the same access, could copy them in weeks.

Investors have learned. A pitch that amounts to "we put a model behind a chat box for a local market" now meets a hard question: what stops someone else doing the same?

What a moat looks like in African AI

• Proprietary or hard-to-get data: local language recordings, agricultural records, health data held under proper consent and regulation
• Distribution: relationships with operators, banks, co-operatives, governments or thousands of small businesses
• Workflow depth: products that do the whole job, not just one AI step
• Local understanding: products that work in local languages, payment systems and regulatory environments that global players handle poorly
• Deep technology: hard engineering, like Satlyt's software for running AI on satellites. AI in orbit

The local advantage is real, if earned

African founders have an edge that global companies struggle to match: knowing how customers actually behave. That M-Pesa matters more than cards, that a sale closes in WhatsApp, that ads must sound like Sheng or Pidgin or Wolof, that power and data are intermittent. Products built around those realities are harder to copy than products built around a model.

Our own answer

At Tari Studio, we think about this constantly. Models are available to anyone. What we build around them is meant to be the hard part: an AI director that writes African scenes, people and languages into every request; brand memory that learns from each team's feedback; scripts, characters and templates; publishing to Facebook, Instagram and TikTok; tracked WhatsApp links into a shared inbox; and payment by M-Pesa. How the Studio works.

Advice for founders

Ask the investor's question before they do: if the model got twice as good and half the price tomorrow, would your business be stronger or weaker? If stronger, you probably have a moat. If weaker, you have a feature.