MTN Group plans to transform thousands of telecom towers across Africa into a distributed artificial intelligence inference network, positioning the operator not only as a connectivity provider but as a foundational infrastructure platform for Africa’s emerging AI economy. The strategy, outlined by MTN Group Chief Technology and Information Officer Charles Molapisi during an event represents one of the clearest signals yet that African telecom operators are beginning to reposition their infrastructure for the AI era.

At the center of the plan is a shift in how telecom towers are architected.

Traditionally, cellular towers rely on baseband units – specialized hardware dedicated solely to operating the radio access network. MTN now intends to replace parts of that infrastructure with open GPU-based configurations capable of supporting both telecom network operations and AI inference workloads simultaneously.

The result, according to Molapisi, would be a continent-wide “distributed AI grid” built directly into MTN’s tower infrastructure.

“The ambition is for MTN to become the biggest distributor of edge inference in the continent,” he said.

The move reflects a broader shift underway globally as operators increasingly view telecom infrastructure not simply as connectivity infrastructure, but as compute infrastructure.

Rather than routing AI workloads back to centralized hyperscale facilities, MTN wants portions of those workloads processed closer to users at the network edge. Molapisi illustrated the concept using gaming latency. In a residential estate served by a local MTN tower, AI-enabled edge compute could process workloads directly at or near the tower instead of sending data to a distant data center and back.

Distributed inference infrastructure could eventually support localized AI applications across financial services, logistics, healthcare, surveillance, manufacturing, smart cities and autonomous systems, particularly in environments where latency, bandwidth efficiency and data sovereignty matter.

The edge AI initiative forms part of a broader infrastructure strategy under the Group’s Ambition 2030 framework, which reorganized the company around connectivity, fintech and digital infrastructure platforms.

Alongside the edge strategy, MTN confirmed earlier this year that it plans to build two AI-enabled data centers in South Africa and Nigeria, creating a dual-layer architecture combining centralized compute facilities with distributed edge inference capacity.

Put together, these initiatives indicate MTN is attempting to build one of Africa’s first integrated AI infrastructure stacks spanning connectivity, fiber, cloud, compute and inference distribution.

MTN’s infrastructure footprint has expanded materially through its growing control and strategic alignment with tower infrastructure across African markets. The company’s evolving relationship with tower operator IHS Towers gives the company access to one of the continent’s largest telecom tower footprints, positioning it uniquely to distribute edge AI infrastructure at scale.

That advantage matters because inference infrastructure depends heavily on physical proximity to users. Unlike centralized AI training clusters, inference workloads benefit from being distributed across dense population and network environments. Africa’s telecom tower networks therefore become strategically valuable not just for mobile connectivity, but as future compute-distribution assets.

By leveraging existing tower infrastructure, power systems and fiber backhaul networks, MTN could potentially accelerate AI infrastructure deployment without needing to build entirely new edge-compute real estate from scratch.

The strategy also creates a potentially important competitive distinction for MTN.

While hyperscalers continue expanding through centralized cloud regions and large-scale campuses, telecom operators possess something many cloud providers do not: deep last-mile physical infrastructure distributed across urban and semi-urban markets. In Africa, where telecom operators often own or influence some of the continent’s most extensive infrastructure footprints, that positioning could become strategically significant in the AI era.

Molapisi said the company’s strategy extends across multiple layers of the AI value chain, including silicon procurement, data center infrastructure, cloud platforms, model ecosystems and application partnerships.

The company is also expanding terrestrial fiber networks across several African markets, including some where it does not operate a mobile network, as part of a wider effort to strengthen what Molapisi described as Africa’s missing digital “rails”.

The announcement also reflects a growing infrastructure sovereignty narrative emerging across African telecom and cloud operators.

Earlier this year, MTN participated in an investment round in US AI-native networking company ORAN Development Company alongside Nvidia, Cisco, Nokia, AT&T and Telecom Italia. At the time, MTN Digital Infrastructure CEO Mazen Mroué framed the move around the principle of “sovereign AI” – the idea that African countries should host and process AI workloads locally rather than relying entirely on offshore infrastructure environments.

The strategy aligns with a broader debate unfolding across Africa’s digital infrastructure ecosystem.

While the continent’s connectivity footprint has expanded rapidly over the past decade through subsea cables, fiber rollout and new data center developments, Africa still accounts for less than 1% of global compute capacity. Industry leaders increasingly argue that the next phase of digital infrastructure competition will not be defined by connectivity alone, but by who controls compute, inference and AI processing environments.

MTN’s tower-based inference strategy may directly address that gap.

By embedding GPU infrastructure directly into telecom networks, the company is effectively attempting to convert Africa’s largest mobile infrastructure footprint into a distributed compute layer.

The economics, however, remain complex.

Molapisi acknowledged that the rapid evolution of AI chips creates procurement risks for operators deploying infrastructure at scale. Nvidia, for example, has already transitioned from Hopper to Blackwell architectures within roughly two years, compressing infrastructure lifecycles and raising the risk of hardware obsolescence.

“If you get that wrong, you’ll get the economics terribly wrong,” he said, referring to the balance between training-focused and inference-focused compute infrastructure.

The announcement marks one of the strongest indications yet that Africa’s telecom sector increasingly sees itself as more than a provider of bandwidth.

In the AI era, telecom towers may no longer function only as radio infrastructure. They may become part of Africa’s future compute fabric.