Okavango AI Cuts Inference Pricing to Undercut Foreign Providers

Okavango AI cut list prices across its hosted inference tier on Sunday night, taking its mid-size model from $0.62 to $0.34 per million input tokens. The company says the cut is permanent and funded by improved utilization on its Gaborone and Cape Town clusters, which it claims now run at 71% average occupancy, up from 44% a year ago.

Price is only half the pitch. Okavango's own benchmarks put median round-trip latency from Lagos at 84 milliseconds against 310 milliseconds to the nearest overseas endpoint most African developers currently use. For chat products that is a nuisance; for voice agents and interactive tooling it is the difference between usable and not.

"We are not going to win on model quality this year and we know it," chief executive Neo Segale told Afrikons. "We are going to win on the boring things — price, distance, and getting invoiced in a currency your finance team recognizes."

The move puts pressure on regional rivals who have been competing on proximity without competing on price. It also raises a question the whole sector will face soon enough: how many independent inference providers a market this size can actually support once the discounting starts.