These two founders left Goldman and Meta to build voice AI for markets everyone else overlooked

The 17,000 daily calls figure is the only hard metric in the story, and it tells us this is already an operational business, not a pitch deck. What the summary skips is the infrastructure angle: building a custom stack for African and Middle Eastern telecom constraints implies the founders aren't wrapping existing APIs, which changes the acquisition and partnership calculus for any larger player eyeing this space.

The technical foundation for this bet got quietly validated just days earlier. The WAXAL-NET paper published June 1st showed that compact, finetuned speech recognition models trained on African languages outperform large multilingual foundation models by 27 percentage points on conversational speech, while running 3 to 40 times smaller. That finding directly supports the thesis that a custom stack beats a generic one in these markets, and it suggests the founders' moat is more durable than it might appear to observers assuming scale wins everywhere. Meanwhile, Alphabet's $80 billion infrastructure raise and SoftBank's $87.3 billion French datacenter commitment both signal that the compute race is concentrating in Western and European corridors, leaving the geographic arbitrage window these founders are exploiting relatively uncrowded for now.

Watch whether a major telecom operator in sub-Saharan Africa or the Gulf signs a distribution partnership with this company within the next 12 months. That would confirm the custom stack is genuinely differentiated rather than a wedge story told to justify the geographic focus.