Technological Transformation
Introduction: Where the Cloud Lives
Nigeria is Africa's largest country by population and one of its most active technology markets, yet its digital economy has long run on infrastructure located elsewhere. Reporting by BusinessDay estimates that Nigeria spends about US$850 million a year on foreign cloud services, and data routed abroad typically faces round-trip latency of 70 to 90 milliseconds, against under 5 milliseconds for a local facility. As artificial intelligence workloads grow, that dependence becomes more costly and, for regulators concerned with data sovereignty, more uncomfortable. Africa holds about 18 percent of the world's population but under 2 percent of global data-centre capacity, and Nigeria is now attempting to close part of that gap.
This article looks at the current state of that attempt: the facilities being built, the money behind them, the GPU access services emerging, and the constraint that threatens to slow everything, which is electricity.
The Numbers: A Market Still Small but Moving
Estimates of Nigeria's current capacity vary by source, which is itself a reminder of how young the market is. TechCabal puts installed capacity at 65 to 86 megawatts and counts 17 operational data centres, while Techpoint Africa cites 86 MW operational, about 25 data centres, and more than 320 MW under construction or planned; TechCabal expects 400-plus MW within three to five years. Techpoint also notes that Nigeria's 30-plus MW of operational IT load represents only around 0.6 percent of global installed capacity, and that South Africa, with 61 data centres, and Kenya, with 19, still lead or match Nigeria in facility counts.
Market-size projections show the same range of outcomes. BusinessDay projects Nigeria's data-centre sector growing from about US$300 million in 2025 to US$800 million by 2031, while Techpoint cites US$374 million in 2026 rising to US$782.8 million in 2031. Whatever the exact baseline, the direction is clear, with more than US$1 billion in committed investment by telecom operators and global infrastructure firms cited by BusinessDay and more than US$2 billion committed through 2027 according to Techpoint.
The Projects: From Enterprise Colocation to AI-Ready Campuses
The most visible shift is from conventional enterprise colocation towards facilities designed for dense AI workloads. Traditional enterprise racks draw 10 to 15 kW, whereas AI-focused racks require 60 to 100 kW and liquid cooling, a specification that few existing Nigerian buildings can meet.
Kasi Cloud's LOS1 campus in Lekki, Lagos, is the flagship. TechCabal reports a US$250 million investment backed by the Nigeria Sovereign Investment Authority, a 4.2-hectare site, a power substation of up to 100 MW described as the largest in Africa, rack densities from 8 kW to 100 kW, and a 95 percent renewable-energy target. Its first 5.5 MW was expected online in the second quarter of 2026, with UduTech as a partner. Airtel Africa's Nxtra unit has committed about US$120 million to a hyperscale facility, with Techpoint reporting a 38 MW build at Eko Atlantic and a 2028 go-live target. MTN's Genova unit opened the first phase of the Sifiso Dabengwa Data Centre in July 2025, a Tier III-compliant 4.5 MW facility, with a second phase adding GPU infrastructure at an estimated US$240 to 250 million and due in the second half of 2026. Equinix is investing US$22 million in its LG3 facility in Lagos, part of a US$390 million five-year pan-African commitment, while Open Access Data Centres operates a 24 MW Lagos site costing about US$240 million, Rack Centre has an AI-ready 12 MW facility, and Zyra Data plans a US$20 million Tier 2 facility in late 2026.
Public and multilateral finance is also moving. Techpoint notes a US$600 million World Bank and EBRD package, Project BRIDGE, to expand fibre from about 35,000 km towards 125,000 km, which would connect data centres to users. Microsoft has committed US$1 million to training in Nigeria as part of a broader Africa initiative, and Google provided a US$1.7 million grant for AI talent development.
GPUs as a Service: Access Before Ownership
Building a data centre is only half the problem; the other half is putting GPUs in reach of Nigerian developers, universities and startups. Cassava Technologies, an NVIDIA cloud partner, launched an NVIDIA-powered AI Factory in South Africa on 19 March 2026, offering GPU-as-a-Service and AI-as-a-Service through local APIs, with Nigeria, Kenya, Egypt and Morocco named as planned expansion markets, though no Nigerian timeline has been published. Locally, Udutech's Africa GPU Hub, launched in August 2025, rents GPU capacity from under US$1 per hour, which lowers the barrier for a founder who cannot afford to buy hardware. This rental model is likely to be the dominant way Nigerian AI builders access compute for the next few years.
For AI developers, local compute matters for three reasons: lower latency for real-time applications, compliance with data-protection rules that favour local processing, and the ability to train and fine-tune models on local languages and datasets without paying dollar-denominated foreign cloud fees during a period of naira volatility.
The Binding Constraint: Power and Imports
Every source that examines Nigeria's data-centre boom arrives at the same warning: power. Techpoint reports that the national grid fluctuates between roughly 5,000 and 6,000 MW against 13,000 MW of installed capacity, and operators who self-generate face costs of US$0.28 to 0.33 per kWh. BusinessDay says diesel backup can account for up to 40 percent of operating costs and that African data-centre power demand is rising 20 to 25 percent a year, potentially reaching 8,000 gigawatt-hours. Roughly 600 million Africans lack reliable electricity, and dense AI racks magnify the problem. Hybrid systems combining gas, solar and batteries are the practical response, and Kasi Cloud's 95 percent renewable target will be an important test of whether it can be met in practice.
The second constraint is import dependence. GPUs, liquid-cooling systems and specialised electrical equipment are all imported, exposing projects to currency swings and supply-chain delay. TechCabal notes typical construction timelines of 16 to 20 months before power connections and commissioning, and that many projects remain in redesign. As a result, it expects the first true AI-focused facility to arrive by 2026, but probably late in the year, and cautions that Nigeria is unlikely to achieve full hyperscale AI capacity overnight.
Assessment: Sovereignty, Skills and Jobs
Data centres are capital-intensive but employ few people once built. TechCabal's estimate for a US$10 million, 1 MW Tier III facility is around 700 construction jobs but only 20 to 30 permanent operating jobs a year. The strongest economic case for Nigeria is therefore less about direct employment and more about the ecosystem effects: cheaper and faster cloud services for banks, fintechs, telecom operators and government agencies; the retention of cloud spend within the country; and a foundation on which AI startups can build.
The next twelve months will show whether the pipeline converts into capacity. Watch for the energisation of Kasi Cloud's first phase, MTN's GPU phase in the second half of 2026, the terms on which Cassava brings its AI Factory to Lagos, and whether reforms to electricity supply for large industrial users keep pace. If they do, Nigeria could become West Africa's AI compute hub; if not, much of its intelligence infrastructure will continue to run offshore.
