Africa's AI startups are building real products, but the economics are tight. A study published on 29 October, AI Made in Africa: Voices from AI Businesses across Africa, finds that most are paying for computing in US dollars while selling to customers who cannot pay much, and that only 15% of the companies surveyed make a profit.
The report was carried out by market research firm Briter for the AI Made in Africa project, which GIZ runs on behalf of Germany's development ministry (BMZ) in partnership with the African Union. It maps 311 for-profit AI companies across the continent and draws on a survey of 85 of them, run between April and July 2026, plus 16 interviews. Most respondents are from Nigeria, Kenya and South Africa.
The compute squeeze
According to the report, 65% of the businesses surveyed say cloud computing and compute are their highest development cost, ahead of collecting and cleaning data. GPUs, cloud storage and model APIs are billed largely in dollars. Technext's review of the study notes estimates that a company serving thousands of daily users through third-party AI APIs can spend $20,000 to $50,000 a month on them.
Prices on the other side are low. The study finds that 56% of the companies sell their products for less than $2,500 a year per customer, about $208 a month. For Nigeria, the report points to the 2023 naira devaluation, which reportedly pushed local cloud bills up by 224%, and to a national grid that averaged about 4,000 MW of available capacity in early 2026 against 13,000 MW installed, according to Technext. Africa is home to about 18% of the world's population but less than 1% of global data-centre capacity.
Related · StartupsAfrican Tech Goes to the UK for Licences, the US for PaperTLP Advisory's study of 100 African tech firms finds most US ties are incorporation only, while UK ties come with licences, teams and customers.Who pays, and who profits
Most of these companies sell to institutions: 88% sell to other businesses and 44% to governments, where sales cycles often run 12 to 18 months. The biggest barrier to growth, cited by 65%, is customer trust and low AI literacy, ahead of high development costs (49%), linking up with older systems (34%) and long sales cycles (33%), Technext reports.
Money is thin. Beyond the 15% that are profitable, 46% earn revenue but are not yet profitable, Ecofin Agency reports, and about a third of respondents are bootstrapped. Technext adds that more than 75% of the $1.3 billion went to companies founded before 2020.
The report includes Nigerian examples. Healthtech startup MyItura built its own clinical datasets and gave grant-funded tablets to primary healthcare workers before doctors would trust its AI tools, according to Technext. Femtech company Venille, which has 3,500 active users, relies on AWS business credits, Ecofin reports.
What founders are asking for
- Subsidies for research, development and computing: 62% of respondents.
- Harmonised data rules across borders, for example between Nigeria and Kenya: 49%.
- Talent: 51% struggle to hire senior AI researchers, and 47% now run their own internal training to upskill junior engineers, according to Technext.
