For years, South Africa’s electricity crisis was one of the biggest obstacles to economic growth. Now Eskom is reportedly talking to some of the world’s largest technology companies about using surplus electricity to power the data centres needed for the next wave of cloud computing and artificial intelligence.
It highlights something often lost amid the excitement around ChatGPT, AI copilots and intelligent agents: AI may exist in software, but the infrastructure that makes it possible is very physical.
“AI is often spoken about as though it happens somewhere in the cloud, disconnected from the physical world. The reality is exactly the opposite. AI requires enormous computing capacity, electricity, connectivity, cooling, data infrastructure and sophisticated engineering,” says Dr Pierre Le Roux, Managing Director of MOYO.
Eskom has reportedly held discussions with technology groups including Amazon, Microsoft and Google as it looks for large new customers for its improved generation capacity.
South Africa already hosts a significant share of Africa’s hyperscale data-centre capacity, while Microsoft has announced a further R5.4 billion investment to expand its cloud and AI infrastructure in the country by the end of 2027.
But Le Roux argues that South Africa should resist viewing this simply as a data-centre investment story.
“The bigger question is whether South Africa can build an ecosystem capable of turning energy, computing capacity, connectivity and engineering expertise into economic value,” says Le Roux.
AI has an electricity problem
Training and running advanced AI models requires significant computing power concentrated in data centres consuming large quantities of electricity. This means access to reliable and affordable energy is becoming an increasingly important part of global technology competition.
“The next phase of the digital economy is going to have a much closer relationship with the energy economy than the last one did,” says Le Roux.
“A country can have great software developers and ambitious AI strategies, but if it cannot provide reliable power, connectivity and computing infrastructure at scale, much of that ambition remains theoretical.”
South Africa therefore enters the AI race with both an advantage and a warning. It has Africa’s most developed data-centre ecosystem, sophisticated financial and telecommunications sectors, established cloud regions, renewable-energy potential and a substantial corporate technology market.
But infrastructure leadership is not permanent.
Infrastructure does not end at the data centre
The same lesson applies inside businesses.
Companies often think about AI infrastructure in terms of servers, cloud platforms and computing capacity. But Le Roux says organisations can have access to the best AI models in the world and still achieve very little if their underlying systems cannot execute.
That includes enterprise platforms, APIs, data architecture, cybersecurity, system integration, governance and, critically, people.
“The AI conversation has become obsessed with the intelligence layer. But intelligence without the ability to act is not transformation,” says Le Roux.
An AI system may identify a maintenance problem, predict a supply-chain disruption or recommend a response to changing customer demand. Creating value requires connecting that intelligence to the systems, processes and people capable of acting on it.
“You need to get from insight to execution. The AI model is only one component,” says Le Roux.
The opportunity is in execution
South Africa’s greatest AI opportunity may therefore not lie in creating the next global foundation model, but in applying AI to industries where the country already has significant experience: mining, energy, agriculture, logistics, financial services, manufacturing and infrastructure management.
“Our opportunity is not necessarily to recreate Silicon Valley in Johannesburg or Cape Town,” says Le Roux.
“It is to take world-class AI capability and combine it with South African engineering knowledge and industry experience to solve real problems. That is where we can build something differentiated.”
Attracting hyperscale data centres would be valuable. But the greater prize is building the capability around them: engineers, data specialists, software developers, integration expertise, energy solutions and businesses capable of applying AI productively.
“That is why the AI race will ultimately be won through execution,” says Le Roux.
“The hype will move on to the next technology. The countries and companies that will still be standing are the ones that built the infrastructure and operating capability underneath it.”
