MTN Group is poised to turn its continent‑wide tower estate into a distributed AI compute fabric, a move that could reshape how South African and African users experience everything from mobile gaming to real‑time analytics. At a Bowmans‑hosted event in Johannesburg, chief technology and information officer Charles Molapisi unveiled a plan to replace the single‑purpose baseband units that sit on every cell site with open‑GPU modules capable of handling both radio traffic and edge‑AI inference workloads. The ambition, he said, is to make MTN the largest distributor of edge and inference capacity across Africa, positioning the telco as the backbone of the continent’s emerging AI economy.
The shift promises dramatic latency gains. Today, many AI tasks – such as image recognition or game‑state calculations – are sent back to centrally located data centres, often hundreds of kilometres away, before a response can be returned to the user. By installing GPU‑enabled compute at the tower itself, MTN can process these workloads locally, shaving milliseconds off round‑trip times and freeing up backhaul bandwidth for other services. Molapisi illustrated the benefit with a simple scenario: a family playing a cloud‑based game on a PlayStation connected through a nearby MTN tower would experience near‑instantaneous response, rather than the lag introduced by a distant data centre.
The edge layer will complement a broader AI infrastructure strategy revealed in MTN’s 2025 financial results, which include the construction of two AI‑enabled data centres – one in South Africa and another in Nigeria. Together, the edge nodes and central facilities form a full‑stack AI ecosystem that spans silicon procurement, data‑centre construction, cloud platform operation, model curation and application co‑development with partners.
MTN edge AI grid: building Africa’s sovereign AI backbone
Molapisi’s presentation painted a picture of a vertically integrated AI strategy that goes beyond connectivity. The group is laying fibre across multiple African markets – even in countries where it does not hold a GSM licence – to create the “rails” needed for high‑speed data movement. This infrastructure push sits within MTN’s Ambition 2030 framework, which reorganises the business around three pillars: connectivity, fintech and digital infrastructure.
Earlier this year, MTN invested in US‑based AI‑native networking start‑up ORAN Development Company, joining forces with Nvidia, Cisco, Nokia, AT&T and Telecom Italia. At the time, MTN Digital Infrastructure CEO Mazen Mroué described the effort as “sovereign AI”, meaning African nations would host the compute that processes their own data, rather than relying on offshore facilities.
Molapisi acknowledged the rapid pace of chip development – Nvidia’s Hopper architecture will be eclipsed by Blackwell within two years – and stressed that MTN is being meticulous about its hardware mix. “If you get the balance between training and inference silicon wrong, the economics collapse,” he warned, underscoring the financial stakes of this transformation.
Current tower hardware vs. future AI‑enabled edge nodes
| Feature | Current baseband unit | Future AI‑enabled edge node |
|---|---|---|
| Primary function | Drive radio access network only | Run radio access and AI inference workloads |
| Hardware type | Single‑purpose ASIC | Open‑GPU platform (e.g., Nvidia A100‑class) |
| Latency impact | Backhaul to central data centre adds 30‑50 ms | Local processing cuts latency to <10 ms |
| Energy consumption | ~150 W per site | ~300 W per site (offset by reduced backhaul) |
| Upgrade cycle | 5‑7 years | 2‑3 years to keep pace with AI chips |
The table shows that swapping out legacy baseband units for GPU‑rich edge nodes could halve latency while introducing higher per‑site power draw – a trade‑off MTN plans to manage through smarter backhaul utilisation and renewable energy sourcing.
The strategic rationale behind the move is clear: Africa currently contributes about 1 % of global compute capacity, a figure that risks turning the continent into a raw‑data exporter while importing the high‑value AI insights it generates. By embedding compute at the network edge, MTN hopes to keep both data and the intelligence derived from it on the continent, fostering home‑grown AI services and reducing dependence on foreign cloud providers.
In practice, the rollout will begin with pilot sites in high‑density urban areas, where the demand for low‑latency AI services – such as augmented reality, telemedicine and autonomous vehicle communication – is strongest. From there, the plan is to scale across MTN’s 140 million subscriber base, leveraging existing tower sites to accelerate deployment and keep capital expenditure in check.
Timeline for the AI edge rollout
| Phase | Key Milestones | Target Completion |
|---|---|---|
| Pilot deployment | Install GPU modules at 50 towers in Gauteng & Western Cape | Q4 2026 |
| Expanded rollout | Reach 500 AI‑enabled towers across South Africa | Q2 2027 |
| Continental scale‑up | Deploy in 1,000 towers across eight African markets | Q4 2028 |
| Full edge grid | All 5,000 MTN towers equipped with AI compute | 2030 (aligned with Ambition 2030) |
The timeline illustrates MTN’s aggressive schedule, aiming to have a functional edge AI grid across the continent well before 2030. Early pilots will focus on proof‑of‑concept use cases, such as real‑time video analytics for smart cities and low‑latency gaming streams.
From a regulatory perspective, MTN is working closely with South African authorities and telecom regulators in other markets to ensure that the introduction of high‑performance compute at cell sites complies with spectrum and security requirements. The company has also pledged to collaborate with local universities and research institutions, creating a pipeline of African AI talent that can develop applications tailored to regional needs.
The edge AI vision also dovetails with MTN’s fintech ambitions. By processing transaction data locally, the network can offer faster, more secure payment verification for mobile money services, a sector that already handles billions of rand in daily volume. Real‑time fraud detection powered by on‑site AI could dramatically improve the ecosystem’s resilience.
Overall, MTN’s push to transform its tower estate into an edge AI grid signals a decisive shift from being merely a connectivity provider to becoming the digital infrastructure backbone of Africa’s AI future. If the rollout proceeds as outlined, South African users could soon experience AI‑driven services that feel instant, secure and locally powered – a stark contrast to today’s model of distant data‑centre processing.
The move also raises competitive questions for other telcos on the continent. As MTN invests heavily in both hardware and software stacks, rivals will need to decide whether to mirror the strategy, double down on traditional network upgrades, or seek partnerships with global cloud players. For now, MTN appears committed to leading the charge, betting that owning the edge will translate into new revenue streams, stronger brand equity and, crucially, a more sovereign African AI landscape.