For artificial intelligence (AI) to thrive, we need a robust and sustainable data centre structure.
Channelwise asked a panel of AI experts about the current state of AI-ready data centre capacity in South Africa and Africa, and what build-out we an expect.
Ziaad Suleman, senior vice-president and CEO at Cassava Technologies, SA and Botswana
Africa currently accounts for less than 1% of global data centre capacity, making a rapid infrastructure build-out an absolute priority if the continent is to lift its share toward 3% to 5% by 2030.
Cassava Technologies is addressing this gap by deploying clustered GPUs within dedicated AI factories currently operational in Cape Town, with Johannesburg set to follow, alongside planned expansions into Nigeria, Kenya, Egypt, and Morocco.
Ravi Bhat, commercial solutions and AI officer at Microsoft South Africa
South Africa is strategically important because AI at scale depends on compute, connectivity, security, compliance and low-latency cloud infrastructure. Through our continued investments, we have positioned South Africa as a key cloud and AI hub and have reiterated ongoing infrastructure investment across South Africa and the wider Global South.
The build-out will continue because demand for sovereign data handling, trusted cloud, secure AI services and local performance is rising. The key question is not headline capacity, but whether the infrastructure can support secure, compliant, high-value AI workloads across industries and public services – in other words, the infrastructure question is not just how much capacity exists today, but whether the right cloud, connectivity and trust foundations are being built for the emerging AI economy.
Othelo Vieira, technical product manager lead at Cloud On Demand
South Africa has the most developed data centre market on the continent, with facilities from operators like Teraco, Africa Data Centres, and the hyperscalers — Microsoft, Amazon, and Google all have a presence. But ‘AI-ready’ is a higher bar than general-purpose data centre capacity. AI workloads require high-density power, specialised GPU infrastructure, and ultra-low-latency networking — and supply of that is constrained. The build-out is accelerating: we’re seeing significant capital commitments from hyperscalers into the South African and broader African markets. Over the next two to three years, we’ll see a meaningful increase in GPU-class compute availability. The constraint that will persist longest is power — reliable, stable electricity is the binding input for AI infrastructure, and that’s a structural challenge for multiple African markets. The channel opportunity here is substantial: helping organisations navigate hybrid architectures, manage costs, and connect on-premises infrastructure to cloud-based AI compute.