Data-centre capacity and energy
Africa accounts for only a tiny number of large-scale data centres. By some estimates, only about 10% of the data-centre demand in Africa is currently served. There is a huge market opportunity for online and connected services, all of which lies on top of data-centre capacity.
Data-centre infrastructure is becoming a global issue. They require enormous amounts of power and water, and deep capital investment. In Africa they can potentially become a force for change, being anchor loads for energy expansion that benefits the countries and communities in which they are located.
The push for data sovereignty and localisation will drive demand higher on a per-country level. Data sovereignty has evolved from an aspiration into a strategic policy lever and will likely continue.
Energy disruption and development challenges
Reducing power disruption across Africa is essential to drive further economic growth. World Bank Enterprise Survey-based analysis found that African firms lose around 5% of annual sales to power outages. The Center for Global Development, using data on more than 3,000 firms in 37 countries, found some firms lose as much as 31% of sales. Costs are also driven up. Backup generation is costly – almost three times the cost of the same energy from the grid.
Standard Bank's Africa Trade Barometer (2024) puts Nigeria's outage losses at $26 billion a year, not counting $22 billion spent on generator fuel. For South Africa, Bloomberg, citing CSIR research, reports that load-shedding cost the economy R2.9 trillion ($169 billion) in 2023, "about a third of GDP", and up to R481 billion ($28 billion) in 2024.
These numbers are a little old and the methodology behind them is uncertain. However they speak to the scale of the problem. There are some green shoots – South Africa has not seen load shedding as frequently – but more capacity is needed.
It is not all about generation capacity, however. EPRA's 2025 Energy and Petroleum Statistics Report puts installed capacity in Kenya at 3,192 MW, against a record peak demand of 2,316 MW in 2025. Once capacity is online, the challenge in Kenya becomes transmission infrastructure.
Investment direction
South Africa's Minister of Communications and Digital Technologies, Solly Malatsi, opened GovTech 2026 in Durban with a speech on the national digital-sovereignty agenda. He described digital sovereignty as a country's ability to make meaningful choices about its data, infrastructure, and online rights without being wholly dependent on third parties.
Malatsi had met officials from Operation Vulindlela in the Presidency to explore positioning South Africa as a global submarine-cable and digital-infrastructure hub, at a time when conflict in the Middle East is disrupting global communications. President Ramaphosa said in SONA 2026 that more than R50 billion of data-centre investment is expected over the next three years.
This push for data localisation will act as a demand lever. The Africa Data Centres Association (ADCA) argues that "clear localisation and data-protection requirements create predictable demand for compliant, in-country infrastructure”.
Hyperscale loads can underwrite new generation where the host system has spare generation and transmission, or where the developer builds captive supply. Where neither applies, as in Kenya, a single campus competes directly with households and industry.
Powering future data centres
Data Center Map counted 243 African data centres in January 2026, nine per cent more than six months earlier, with 60 in South Africa and 22 in Nigeria.
There are a range of power models available for new data centres. In South Africa, operators combine Eskom supply with wheeled renewables and on-site solar. Rack Centre's 13.5 MW off-grid gas plant and Tetracore's planned 100 MW on-site gas plant represent the Nigerian model of bypassing the grid entirely. Kenya has delivered a geothermal approach which offers firm, low-carbon power, but the Olkaria project shows that resource availability does not resolve transmission, redundancy and fiscal constraints.
Africa still has a major renewables advantage. There are abundant solar, wind, hydro, and geothermal opportunities; however, diesel backup is still the universal backup strategy across many communities and businesses.
A major driver for data-centre capacity is AI and the value it may create. The IMF (July 2026) estimates AI could make Sub-Saharan Africa's economy around 4% larger over a decade, but it is conditional on electricity, internet access and skills.
Challenges of concentration
Africa had roughly 360 MW of active data-centre capacity at the start of 2026 (Africa Data Centres Association), about 0.6% of global capacity, and the IEA projects Africa's data-centre consumption per head will remain under 2 kWh in 2030. The problem is not continental energy balance; it is that single campuses of 100 MW to 1 GW land on national systems of 3 to 5 GW.
Regulatory choices
One focus area is accelerating AU Member States' capabilities in energy infrastructure, broadband, computing, and data centres.
Data Centres for AI usage require advanced hardware, much of which is of US origin. US concerns about Chinese AI dominance have led to restrictions on external supply. US GPU access is now tied to "full-stack" American packages. How do African governments preserve genuine choice in their AI infrastructure?
We see a small number of key policy decisions:
- open, regulated third-party access and wheeling regimes;
- transparent large-load connection rules that allocate grid-reinforcement costs to the connecting customer;
- sequencing localisation mandates against verified in-country capacity;
- water-use disclosure for cooling
- designation of data centres and grid control systems as critical infrastructure with enforceable security standards.
Sharing cost and risk
These regulatory changes need to be supported with a clear view of who bears risk, and how it can be shared to drive investment.
It is likely that initial investment will come from Hyperscaler, the largest-scale cloud computer and data centre providers, and their balance-sheets. They are well resourced and can offset ramp-up costs in Africa against excess demand elsewhere in the world.
Providing power infrastructure will need a different model to relying on these providers. The hyperscalers market and data centre customers expect power and cooling to be a solved problem to some degree, which is based on their experience in their home countries with highly developed infrastructure.
There will be a need for transmission risk-sharing, where developers, governments and utilities providers share the financial or delivery uncertainties of accessing and enhancing high-voltage power-lines. South Africa's ITP programme and Credit Guarantee Vehicle are the most developed African attempt to move transmission risk off the sovereign balance sheet, and it may signal a model others can follow.
There will be a risk of emerging US export controls in relation to the technology required to support AI infrastructure. The Biden administration's AI Diffusion Rule would have imposed a worldwide licence requirement on advanced chips and tiered country access. These risks are only likely to increase in the current geopolitical climate.
It is still not clear who bears the stranded-asset risk, where data centres values are not supported by the market or the inherent currency risks of international expansion. Data centres are not filled overnight, and there is a risk that demand takes time to build beyond what initial project valuations may expect. Many of the obvious financial backers will be concerned more about currency risks, given that building and running costs are denominated in local currency, while investors’ balance sheets are in hard currency.
There is a wider backdrop of physical security and resilience needed. Data Centres share much similarity with oil and gas sites, where there is a well-trodden security path. They differ however in that the impacts of an incident are felt far beyond the location of the data centre, and the data centre itself is heavily reliant on country and continent level connectivity.
Major continental connectivity is via subsea cables. These have been thrown into sharp focus given their targeting by major countries such as Russia. Aside from deliberate damage or misuse, subsea cables can be vulnerable when they are near the seabed surface. For example, Recorded Future's Insikt Group logged 44 cable damage events across 32 locations between 2024 and mid-2025, around a quarter linked to anchor dragging.