The World Bank, in partnership with Cassava Technologies and Nvidia, announced a programme that will provide pay‑as‑you‑go access to GPU computing for governments and businesses in emerging markets, with an initial focus on African countries.

The initiative is presented as an alternative to each country building its own national data centre, a process described as costly and slow to establish.

Through a common system, participating institutions can obtain specialised hardware without the capital outlay of constructing large facilities, potentially making AI computing more practical for organisations that lack the resources to build dedicated data centres.

The service is part of a broader discussion on AI infrastructure that featured at the Moonshot 2026 conference, where AI and machine learning were highlighted as major themes.

While shared GPU access can lower entry barriers, the availability of local data centres remains crucial; recent expansions such as MTN’s Africa Data Hub and Digital Realty’s NBO2 facility in Kenya add capacity for cloud and AI services.

Energy demand is a key consideration, as AI workloads consume substantial electricity; projects like AMACO’s proposed LNG‑powered AI data centre in Mombasa illustrate the link between computing infrastructure and power planning.

For African startups and technology firms, access to shared GPU resources could reduce the cost of training models and running AI applications, influencing product pricing and market reach.

Public sector entities exploring AI for research, data processing or service delivery also stand to benefit, as the shared model offers a way to obtain specialised hardware without building a national facility.