MTN Group plans to transform its African tower network into a distributed AI computing fabric by installing open GPU infrastructure at base-station sites, enabling the same hardware to support both its mobile network and edge-AI inference workloads.
Charles Molapisi, chief technology and information officer of MTN Group, presented the strategy at a gathering in Johannesburg organised by the legal firm Bowmans. It was MTN’s most thorough description to date of how it intends to position itself as Africa’s AI economy’s infrastructure layer.
A baseband unit, a piece of specialised gear used solely to power the radio access network, is now found at the base of every cellular tower. According to Molapisi, MTN will swap them out for open GPU setups that can handle the radio together with AI inference, or what the business has called a “distributed AI grid.”
He argued that latency is a payoff. AI workloads could be processed at or close to the tower instead of being transported back to a central data centre. He used the example of kids playing PlayStation on a tower-served estate to illustrate how installing edge compute would allow the workload to be serviced locally rather than being backhauled to a remote data centre and then returned, freeing up capacity and reducing round-trip time.
The centralised portion of MTN’s AI infrastructure strategy is positioned next to the edge layer. The company announced in its March 2025 financial results that it would construct two new AI-enabled data centres, one in Nigeria and one in South Africa.
Edge Artificial Intelligence Grid
Molapisi outlined an MTN AI strategy that encompasses a comparatively comprehensive stack, including acquiring silicon, constructing data centres, managing its own cloud platforms, selecting models and collaborating with partners to develop applications.
In order to fill what Molapisi referred to as the continent’s “rails,” the company is also constructing terrestrial fibre throughout several African markets, even some where it lacks a GSM licence.
The investments are part of MTN’s Ambition 2030 plan, which reorganised the company around three platforms: digital infrastructure, fintech and connectivity.
The tower-to-inference drive is the most tangible expression of a theory that MTN has been outlining for over a year.
In March, the company invested alongside Nvidia, Cisco, Nokia, AT&T and Telecom Italia in the US AI-native networking start-up ORAN Development Company. The movement surrounding “sovereign AI”—the idea that African nations could host AI computation domestically rather than depending on offshore infrastructure—was explained at the time by Mazen Mroué, CEO of MTN Digital Infrastructure.
According to Molapisi, MTN is collaborating with tech partners on the edge AI grid. He stated that MTN wants to be “the biggest distributor of edge inference in the continent.”
Molapisi’s broader argument that Africa runs the risk of repeating its commodities history in the AI era forms the basis of the strategic thesis. According to him, Africa has the potential to “export raw data” in the same manner that it has long exported raw resources, only to import the intelligence derived from it at a premium, as the continent currently has approximately 1% of the world’s computing capacity.
In his talk, Molapisi acknowledged that procurement decisions made now may become outdated by deployment due to the rapid turnover of chip generations, such as Nvidia’s Hopper to Blackwell in just two years. Because “if you get that wrong, you will get the economics terribly wrong,” he explained, MTN is careful about its chip mix and the ratio of training to inference silicon.
