MTN Group CEO, Ralph Mupita, announced at a press roundtable on Tuesday that the group aims to provide 150 MW of capacity as part of the first phase of developing artificial intelligence data centers. This initial phase will cover South Africa and Nigeria. He specified that the group will hold only a minority stake in the data center enterprise, which is supported by the UAE.
These statements provided concrete figures regarding a deal that MTN had previously described only in general terms in its six-month results report ending June 30, 2026. The report confirmed that after the reporting period, MTN Digital Infrastructure entered into a strategic partnership with a UAE-backed data center investment platform to accelerate the creation of AI-ready digital infrastructure platforms across Africa.
This entity, named Africa Data Hub Holding, is intended for developing and scaling future digital infrastructure capabilities in key African markets. However, the initial MTN report did not disclose its stake, planned capacity, or invested capital, nor did it name the partner. Ralph Mupita provided some of this missing information on Tuesday.
He emphasized that the company's approach is to collaborate with third parties possessing the necessary skills and capabilities, similar to the mentioned UAE-backed deal, and that MTN will act as a minority investor. He added: 'Currently, we are in this partnership only in South Africa and Nigeria, considering 150 MW as the first phase. We will expand as demand arises. We are securing land acquisition and necessary power agreements for the construction of these data centers in several phases as needed.'
Corporate Sector and Cloud Technologies
The minority investor position aligns with the capital-efficient approach that MTN signaled by attracting partners to implement the AI data center project in Africa, rather than financing high-scale capacity from its own balance sheet. This also corresponds with the repositioning within the group's Ambition 2030 strategy, which emphasizes partnerships in fintech and digital infrastructure while the group maintains strict control over debt levels. During the reporting period, MTN achieved a net debt to EBITDA ratio of 0.3x and did not announce an interim dividend, but confirmed the start of a share buyback worth 6 billion Rand.
The demand driving this push for data center development comes from the corporate sector and cloud services, not the retail mobile market. MTN states that the platform will meet the 'growing demand for cloud, enterprise, and resource-intensive services' across Africa, expecting the continent's corporate market to grow by approximately 1.6 times by 2030.
Mupita noted that these clients and workloads differ from those served by the connectivity business, which still generates the majority of revenue. Meanwhile, revenue from group services grew by 17.5% in constant currency, reaching 115.3 billion Rand for the half-year. Previously, MTN announced plans to transform its African telecommunications assets into an AI inference network.
When asked about the most interesting AI use cases, Mupita did not mention consumer products but focused on the network itself. He stated: 'When using AI, optimizing energy efficiency in our networks is an obvious step. We are already considering Cape Town as a test base for dynamic energy consumption reduction. So it might sound boring, but network efficiency is what interests me most right now.'
Energy is one of MTN's largest operating expenses, and the group stated that its first AI goal is the company itself, applying this technology in internal operations before offering it to others.
Model Routing
Regarding advanced AI models, Mupita stated that MTN is less interested in supporting a single winner than in creating a system capable of routing tasks to the model that performs best. He noted: 'Frontier AI models and how we integrate them is an ongoing question, especially for the corporate sector. It is a rapidly evolving area that surprises us every two or three months.' He continued: 'We are exploring how to create an open architectural system that combines different types of models, such as frontier models combined with open-weight models from China, to automatically find the most optimal model for a given task. This is one of the approaches we are considering.'
Furthermore, Mupita has joined the new Global AI Committee, continuing the work of the Broadband Commission. He was appointed to this committee earlier this year. He shared his impressions: 'It was an interesting discussion, including issues on how to ensure that the AI race does not lead to a digital underclass. It is also a geopolitical issue—ensuring access for the Global South. And ensuring that companies and AI players act ethically.'
2