African mobile operators build AI models in local languages
A new GSMA report reveals that Africa's top mobile operators, including MTN, are actively developing AI language models trained on African languages, representing a major infrastructure play in the AI stack. This positions telecoms as key AI infrastructure providers rather than passive connectivity pipes, with implications for data sovereignty and local AI product development.
African AI: the data is the infrastructure
The Swahili reasoning model demonstrated at MWC Barcelona is interesting. What sits underneath it is more interesting.
Google's WAXAL dataset — three years of work with African universities, including the University of Ghana — captures speech across 27 languages spoken by more than 100 million people, and it's open for anyone to build on. That's the detail worth pausing on.
For Ghanaian developers, this is the layer that was always missing: not ideas, not ambition, but the raw material to train something real. The University of Ghana helped build it. Now builders in Accra can use it.