SA’s data centre power demand is expected to almost double over the next five years, as artificial intelligence (AI) drives demand for computing capacity, but incorrect forecasting could result in power outages, US experience suggests.
Installed IT load capacity is forecast to increase from 435MW in 2024 to 829MW by 2029, a compound annual growth rate of 17.5%, according to the Development Bank of Southern Africa’s (DBSA’s) South Africa’s Digital Infrastructure Investment Study 2025. That is almost the entire generating capacity of one of Koeberg’s two reactor units.
A typical AI-focused data centre consumes as much electricity as 100 000 households, but the largest ones under construction today will consume 20 times as much, according to the International Energy Agency (IEA).
US experiences recounted by Stephen Bessasparis, senior consultant at Energy + Environmental Economics, writing for the Bulletin of the Atomic Scientists, suggest getting that demand forecast wrong could have real consequences − from lost investment to power outages − if the grid is pulled harder than expected.
Building for growth
DBSA’s study attributes the increased demand to organisations migrating workloads to the public cloud, growing adoption of digital services, increased use of AI technologies, and rising demand for data storage and processing.
To support this growth, the study says SA will need infrastructure beyond its existing hyperscale facilities in Johannesburg, Cape Town and Durban. It estimates the country will need four to five new regional data centres, each with around 1MW of IT capacity – enough to support about 200 server racks – plus 20 to 30 edge data centres in underserved areas over the next five years.
The study also notes that investment in local data centre infrastructure is expected to rise from more than R39 billion in 2023 to R68 billion by 2029, with global hyperscale operators, including Amazon Web Services, Google, Huawei and Microsoft, investing alongside local operators.
As electricity demand grows, operators are increasingly turning to renewable energy to supplement grid supply and meet sustainability targets, the DBSA says. Many facilities are investing in renewable energy projects through Eskom’s wheeling arrangements, which allow electricity generated elsewhere to be supplied through the grid to their data centres, the report notes.
Global picture
Internationally, electricity demand from data centres rose 17% in 2025, almost matching the growth rate DBSA forecasts for SA’s own capacity, against just 3% growth in overall global electricity demand, driven by continued investment in AI infrastructure, according to the IEA’s Key Questions on Energy and AI report, released in April.
The report says data centre electricity consumption is expected to double by 2030, while power demand from AI-focused data centres is forecast to triple.
AI systems are becoming more efficient and using less electricity per task, the IEA says, but overall power consumption continues to rise, driven by rapid AI adoption and the growing use of more energy-intensive applications, such as AI agents.
Powerful forecasts
Yet accurately forecasting how much electricity AI-driven data centres will ultimately require remains one of the biggest challenges facing the energy sector.
Bessasparis points to 1983, when utilities secured funding through bonds, expecting power demand to continue growing at 7% a year, which did not happen. The Washington Public Power System defaulted on $2.25 billion – worth roughly the same in rand at the time, when the two currencies traded close to parity.
This is still the largest public bond failure in US history and resulted in investors carrying the cost, with the value of their investment dropping 90%, writes Bessasparis.
“The bond disaster was, in part, due to the utility company’s inability to predict the future… For example, to meet the energy needs of 2040, utilities need to begin planning today. Load forecasts are a critical part of those processes,” says Bessasparis.
Locally, Eskom too is funded partly through bonds, alongside the money it earns from selling electricity, government financial support and loans from international development institutions.
Unplugged disaster
Under-forecasting can also lead to serious issues, says Bessasparis, highlighting a summer 2020 issue when California utilities had to unplug nearly half a million people because of “unprecedented heat waves”.
Data centres will be the “most important driver of nationwide electricity demand, and inaccurate forecasts could lead to disruptive over- or under-investments in electricity infrastructure,” says Bessasparis.
“Facilities are being built in anticipation of future artificial intelligence computing needs. The true pace, and ultimate adoption level, of AI is unknown.” He adds that AI could be anywhere on Gartner’s Hype Cycle.
“AI poses a foundational challenge to our electric grid. By understanding the when, where and how much, people can chart a path towards the best future for themselves, their society and the world. Humans are not powerless.”
Share
