AI Divide in Africa: 38% internet, <1% data centers, paper warns

A new arXiv paper by Abayomi O. Agbeyangi and Jose M. Lukose provides a data-driven analysis of the AI divide in Africa, focusing on three dimensions: infrastructure, accessibility, and human capacity. The study reveals stark physical constraints: only 38% internet penetration across the continent, poor broadband coverage, and fewer than 1% of global data centers located in Africa. These infrastructure gaps are compounded by high data costs relative to local incomes, which limit both individual access and organizational scaling.

The paper also examines human-centered factors, including a gender-based digital divide and the need for natural language processing (NLP) models that can understand Africa's diverse native languages—most current systems are trained on English and other high-resource languages. Despite these challenges, the authors identify positive grassroots movements: local startups and universities are increasingly contributing to AI research and development. They offer concrete policy recommendations—such as investing in data center infrastructure, subsidizing connectivity, and funding language-specific AI datasets—to help close the divide and foster an inclusive AI ecosystem.