Cassava’s AI factory and the World Bank’s plan for shared African computing power
Training and running modern AI requires specialised chips called GPUs, which are expensive and scarce in Africa. A Zimbabwean-founded company is trying to change that.
The AI factory
Cassava Technologies, part of the group founded by Zimbabwean billionaire Strive Masiyiwa, began deploying what it calls an “AI factory” at its South African data centres in 2026, in partnership with US chipmaker Nvidia.
The facility offers “GPU-as-a-service” and “AI-as-a-service”, allowing local businesses, researchers and developers to rent computing power instead of buying their own expensive hardware. When the partnership was announced in 2025, it was reported to be worth about $720 million. Masiyiwa said in November 2025 that the company had acquired 12,000 Nvidia GPUs.
In October, Cassava added Nvidia’s Nemotron open AI models to the AI Factory as a ready-to-use service. Businesses can call them through its CAIMEx platform without running their own model infrastructure.
Cassava plans to expand to Nigeria, Kenya, Egypt and Morocco.
“Our goal is to give Africa the infrastructure to write its own future, using its own languages,” said Ahmed El Beheiry, Cassava’s group chief technology and AI officer.
The World Bank steps in
In October, the World Bank Group announced an AI Infrastructure and Readiness Initiative alongside its first digital strategy. The idea is to give governments in emerging markets shared, pay-as-you-go access to GPUs, rather than each building costly national data centres. Africa and Latin America are the initial focus, and Cassava and Nvidia are named as partners.
World Bank President Ajay Banga said the focus would be on “small AI”: practical tools built for specific tasks and local conditions.
Why it matters
- Cost. Few African universities or start-ups can afford their own GPU clusters.
- Data sovereignty. Processing data in Africa, rather than abroad, can help comply with data protection laws and keep sensitive information at home.
- Languages. Local computing power makes it easier to build AI that works in African languages.
The questions
- Electricity. AI data centres need large amounts of reliable power, which remains a problem in many African countries, including South Africa.
- Affordability. Rental prices will determine whether small users can actually benefit.
- Concentration. Relying on a few private providers for critical infrastructure raises questions about competition and control.
What to watch
- Launch dates in Nigeria, Kenya, Egypt and Morocco.
- Which governments join the World Bank initiative.
- Pricing and access for universities and start-ups.
Sources: Billionaires.Africa, 20 March 2026; TechTrends Kenya, 20 October 2026; Bloomberg; Connecting Africa; Techweez, 21 October.
Top photo: Johannesburg’s central business district (file photo). Photo: Evan Bench, Wikimedia Commons, CC BY 2.0
