MTN, Airtel and Orange build African-language AI models via GSMA
A new GSMA report details how Africa's three largest mobile operators — MTN, Airtel, and Orange — are developing large language models trained on African languages. The initiative represents the continent's most coordinated telecoms-led AI infrastructure effort to date.
G6 Languages: the telcos have something no AI lab does — 300 million existing customers who speak Swahili, Amharic, and Hausa
The reason African languages get left out of mainstream AI isn't malice — it's data. English dominates because there's more English text to train on. That's a structural problem, and no startup can fix it alone.
What the G6 operators bring is scale nobody else has: billing records, voice calls, SMS threads, and direct relationships with hundreds of millions of people whose languages have never appeared in a training dataset. Google's WAXAL project took three years and multiple African universities to cover 27 languages. The telcos can go wider, faster, because the raw material is already in their networks.
The working Swahili reasoning model shown at Barcelona is small proof that this is moving. But the real question is whether the operators treat this as shared infrastructure — or as a competitive moat they each build separately.
Story timeline · 2 days
- Jun 23, 20267MTN repurposes cable infrastructure for AI sensing
MTN's AI infrastructure activity has expanded from African-language model development to repurposing physical cable assets for AI sensing applications.
- Jun 18, 20268· this storyMTN, Airtel and Orange build African-language AI models via GSMA