In Ikeja, on the outskirts of Lagos, workers are preparing for the second and most consequential phase of what will become Nigeria’s largest data centre. The Sifiso Dabengwa Data Centre — named after MTN’s late former CEO — opened its first 4.5-megawatt phase in July 2025, housed in 96 containerised modules across a Tier III-compliant facility with 1,500 server racks. Phase two, expected in the second half of 2026, will add AI-optimised GPU infrastructure and double the facility’s capacity to nine megawatts. Total investment: $240 million.
In Olkaria, south-west Kenya, a geothermal field best known for powering the Rift Valley’s grid is about to host something different: a 100-megawatt data centre campus funded by Microsoft and Abu Dhabi’s G42, at a cost of $1 billion. The facility will provide the physical backbone for Azure’s East Africa Cloud Region, expected to become operational around May 2026. Eventually, if demand justifies it, that 100-megawatt initial capacity could rise to one gigawatt.
And in Johannesburg, Google Cloud’s South African region — live since January 2024 — is being threaded into a wider continental infrastructure strategy anchored by two subsea cable systems: Equiano, which already connects Portugal to South Africa via West Africa, and Umoja, announced for service in 2027, which will link Australia to Africa and extend terrestrially through East Africa from South Africa to Kenya.
For the first time, Africa’s data centre investment wave looks serious. Whether it will be enough is another question entirely.
The Scale of the Gap
The African Data Centre Association’s 2026 economic report, published in February, puts the continent’s active data centre capacity at 360 megawatts. That is against a global active capacity of 5,500 megawatts. Africa’s share: approximately 0.6 per cent of the world’s installed compute infrastructure, for a continent that is home to 18 per cent of the world’s population and some of its fastest-growing digital economies.
The pipeline is more encouraging: 238 megawatts under construction and 656 megawatts planned across the continent. If all announced projects materialise, Africa could approach 1.2 gigawatts of total capacity. But the ADCA report carries a sobering caveat: even at that level, Africa is projected to maintain — not increase — its 0.6 per cent global share, as hyperscale data centre construction accelerates in the United States, Europe and Asia at a pace that dwarfs anything available in Africa. Global hyperscaler capex is projected at $602 billion in 2026. Amazon, Microsoft, Google and Meta are each exceeding $100 billion individually. Africa’s entire pipeline represents less than three days of global hyperscaler spend.
The market numbers are more optimistic. Africa’s data centre sector was valued at $1.94 billion in 2025 and is projected to reach $4.36 billion by 2031, at a compound annual growth rate of 14.46 per cent. Nigeria’s sub-sector alone is expected to grow from $300 million in 2025 to nearly $800 million by 2031. That trajectory is real — driven by rising cloud adoption, digital financial services, and increasing AI workloads from the continent’s growing developer community.
The problem is that demand is growing faster than supply, and supply is constrained not by capital — capital is arriving — but by energy.
MTN’s Infrastructure Bet
MTN’s data centre push sits within the Ambition 2030 strategy, which reorganises the group’s business around three pillars: connectivity, fintech, and digital infrastructure. The infrastructure pillar has three sub-priorities: fibre, AI-enabled data centres, and tower value. Data centres are not a side project — they are one of three strategic bets the continent’s largest telecoms group is making for the next decade.
The Sifiso Dabengwa Data Centre is developed under MTN’s Genova infrastructure unit. Phase one, commissioned on 1 July 2025, brought the group’s first major colocation and cloud facility to Nigeria — a market MTN knows intimately from its 27 million mobile subscribers. The facility offers hyperscaler-adjacent connectivity, Tier III resilience, and local sovereign cloud capability for Nigerian enterprises and government entities that have increasingly specific data-residency requirements following the Nigeria Data Protection Act.
Phase two changes the story’s scale. AI-optimised GPU infrastructure — the hardware stack that underpins generative AI inference, model fine-tuning, and vector database operations — will be installed in the second half of 2026. MTN has not named the GPU partner, but has confirmed it is in negotiation with strategic partners on co-investment structures and operating models. The group plans to provide detailed capital allocation information at a Capital Markets Day scheduled for June 2026.
South Africa is the group’s second priority market for the data centre strategy, with separate greenfield development under assessment. Both Nigeria and South Africa were confirmed as priority markets following a comprehensive market study completed in the first quarter of 2025.
Microsoft’s Geothermal Gamble in Kenya
Microsoft’s entry into Africa’s data centre market is characteristically ambitious — and it has chosen an unconventional power source to try to avoid the energy problem that bedevils Nigerian facilities.
The partnership with G42, the Abu Dhabi AI specialist, was announced in May 2024. The $1 billion initiative was structured as a comprehensive digital ecosystem play: not just infrastructure, but East Africa Cloud Region access for Azure customers, Swahili and English AI model development, a 100,000-person AI skills training programme, and a commitment to extend broadband connectivity to 20 million Kenyans by the end of 2025.
At the centre of it is the Olkaria data centre campus. Geothermal energy — drawn from heat stored beneath the Rift Valley — provides the power source. Kenya already generates approximately 43 per cent of its electricity from geothermal sources, giving Olkaria facilities a cleaner, more stable power supply than any diesel-dependent alternative. Initial capacity of 100 megawatts is expected at commissioning, with a long-term potential of one gigawatt if demand warrants the build-out.
The Azure East Africa Cloud Region gives Microsoft its first African cloud region, positioning it alongside Google (Johannesburg, live since January 2024) in terms of regional cloud presence on the continent. For developers in Nairobi, Kampala, and Dar es Salaam, it means Azure services running from infrastructure physically located in East Africa for the first time — with implications for latency, data residency compliance, and the effective cost of AI compute workloads.
The Energy Paradox
Nigeria is Africa’s largest economy and its most active digital market. It is also the continent’s most instructive cautionary tale for data centre operators.
Data centre facilities in Nigeria routinely budget diesel backup for weeks-long power gaps. Grid instability elevates energy costs to 55 to 65 per cent of operating expense at Nigerian facilities, compared with 35 to 45 per cent in mature markets. That 10-to-20-percentage-point gap compounds across the life of a facility: a data centre that would generate 15 per cent margins in Singapore or Amsterdam can generate 2 to 5 per cent in Lagos, before any consideration of generator maintenance costs, fuel theft, and logistics.
The operators adapting most effectively are embedding renewable energy solutions into site design from the beginning, rather than treating diesel as the primary backup. Solar, battery storage, and hybrid power systems are increasingly standard in new facility announcements. Kasi Cloud’s planned 100-megawatt AI campus in Lekki — with 5.5 megawatts expected online in the second quarter of 2026 — incorporates on-site renewables into its power architecture from day one. MTN’s Phase two design for Sifiso Dabengwa includes power resilience planning, though the group has not disclosed specifics.
Kenya’s geothermal advantage is real but not easily replicated. South Africa has adequate grid infrastructure in its data centre corridors. Nigeria, Ghana, and much of West Africa do not. The practical implication: the most aggressive AI infrastructure investments on the continent are concentrating in markets where the energy problem is either solved (Kenya’s geothermal) or manageable (South Africa’s Eskom-adjacent corridors), while the continent’s largest digital economy — Nigeria — remains constrained by the grid.
| Project | Location | Capacity | Investment | Timeline | Power Source |
|---|---|---|---|---|---|
| MTN Sifiso Dabengwa DC — Phase 1 | Ikeja, Lagos, Nigeria | 4.5 MW (Tier III) | $120M | Operational July 2025 | Grid + diesel backup |
| MTN Sifiso Dabengwa DC — Phase 2 (AI GPU) | Ikeja, Lagos, Nigeria | +4.5 MW (AI-optimised) | $240M total | H2 2026 | Grid + diesel backup |
| Microsoft + G42 East Africa Data Centre | Olkaria, Kenya | 100 MW (up to 1 GW) | $1B | ~May 2026 | Geothermal |
| Google Cloud South Africa Region | Johannesburg, South Africa | Undisclosed | Undisclosed | Live (Jan 2024) | Grid |
| Kasi Cloud AI Campus (Phase 1) | Lekki, Lagos, Nigeria | 5.5 MW (of 100 MW) | Undisclosed | Q2 2026 | Hybrid (solar + grid) |
What This Investment Wave Actually Means for African AI
The data centre investment wave solves one part of Africa’s AI compute problem: proximity. If MTN’s Phase two AI GPU infrastructure comes online in H2 2026 as planned, a Lagos developer will, for the first time, be able to rent GPU compute from a facility physically located in Nigeria — potentially satisfying data-residency requirements, eliminating cross-Atlantic latency, and paying in a billing arrangement that does not require converting naira to dollars at a punishing exchange rate.
The Microsoft-G42 Azure East Africa Cloud Region is more significant still. Azure’s GPU families — including the A100-class instances used for serious model training — have been unavailable in any African region. An East Africa Cloud Region with AI infrastructure could give Nairobi’s developer community access to training-grade compute at sub-100-millisecond round-trip times for the first time.
What these investments do not solve is pricing. Hyperscaler on-demand GPU pricing does not vary by region in ways that accommodate African income realities. An A100 instance costs the same in Nairobi as in Amsterdam. The currency risk, access-to-credit asymmetry, and structural cost disadvantage that Africa’s AI developers face against their US and European peers does not disappear because the data centre is now located in the same time zone.
The ADCA projects Africa’s data centre market reaching $4.36 billion by 2031. That trajectory, if it holds, will represent genuine infrastructure maturation. But the continent will need approximately ten times its current 360-megawatt active capacity to meaningfully shift its 0.6 per cent global share — and even the optimistic pipeline scenarios do not reach that figure by 2030.
The race into Africa’s server rooms is real. The scale required to close the gap is, for now, larger than the race itself.
Disclosure
BETAR.africa has no commercial relationship with MTN Group, Microsoft, G42, Google, Kasi Cloud, or any organisation named in this report. All investment figures, capacity data, and market projections are sourced from public company statements, the ADCA 2026 Economic Report, and third-party analyst data cited in the text.
