What it knows

Currently at 14 facilities across 5 countries - South Africa, Kenya, Nigeria, Egypt, and Morocco - covering hyperscaler cloud regions, large carrier-neutral colocation campuses, and purpose-built AI compute facilities announced, under construction, or expanded within the last three years.

What the dataset shows so far: African AI compute is heavily concentrated in South Africa, and most of the headline capacity is announced rather than operational.

How it is built

A data pipeline where sources are triaged in an inbox, promoted to cited evidence, and only then written into the dataset - one YAML file per facility, validated against a JSON schema at build time.

Every data point carries a source, a retrieval date, and a trust tier: company filings and official announcements are weighted above trade press, local journalism, and social media, and the tier travels with the value onto the page, so a reader sees how much weight a given number has. The dataset is version-controlled and openly licensed (CC-BY 4.0).