By Faith Waithaka, cloud and service provider segment lead for Anglophone Africa at Schneider Electric.
Artificial Intelligence (AI) is no longer a futuristic concept; it is rapidly becoming the backbone of modern business and society. Yet the conversation cannot stop at ChatGPT and other chat-based tools. The next wave of innovation lies in agentic AI and fully integrated applications, systems capable of autonomous decision-making and seamless orchestration across industries.
For Africa’s data centres, this shift demands urgent attention: the infrastructure that powers today’s AI will not be sufficient for tomorrow’s AI. Preparing for this future is not simply a technical challenge, but a strategic imperative for the continent’s competitiveness and resilience.
To unpack how AI has evolved from ChatGPT to agentic AI to AI integration with business applications, it is useful to start with the differences between conventional data centres and those designed for AI. Traditional servers rely on standard Central Processing Units (CPUs), which handle one calculation at a time. By contrast, AI workloads demand high‑compute CPUs and Graphics Processing Units (GPUs), which are specialised processors capable of performing billions of operations simultaneously.
Understanding GPUs
GPUs are often misunderstood. While originally designed for graphics, their real strength lies in speed and parallel processing. They can pull data from multiple sources simultaneously, learn patterns, and generate outputs, whether accurate or flawed, depending on how they are trained.
This distinction matters because GPUs deliver far greater capacity than CPUs, but at a cost: they consume significantly more power and generate more heat, requiring advanced cooling systems. In short, the infrastructure behind AI is not just faster; it is heavier, hotter and more demanding. Recognising this is the first step towards understanding why tomorrow’s AI data centres must be built differently from today’s.
Unfortunately, African data centres face a daunting challenge as they prepare for agentic AI, with the most immediate issue being power. Kenya’s grid, for instance, operates at 78-84% of its 2.3GW capacity, yet a single AI training facility could consume close to 1GW alone. Current surplus generation will not be enough, so Africa must plan for exponential demand. Connectivity also lags behind the needs of real-time AI, with fibre penetration and regional interconnectivity still uneven.
Nurturing Africa’s talent
Cooling is another hurdle as high-compute CPUs and GPUs generate far more heat than conventional servers, requiring advanced solutions such as liquid or immersion cooling to remain sustainable. Liquid cooling may sound complex, but at its core, it is simple: cold water is pumped close to the chip, heat is carried away through a manifold and released into a chiller system. This allows GPUs and CPUs to run billions of calculations without overheating.
Perhaps the greatest opportunity (and risk) lies in talent. Africa’s youthful population could be a powerful driver of AI innovation, but only if equipped to build models, integrate them into business applications and manage the infrastructure. This demands deliberate investment in training and collaboration across government, industry and academia.
Demystifying AI infrastructure
This is where scalable, tested configurations such as Schneider Electric’s 132-kilowatt AI system become vital. In such a setup, racks of servers are powered and cooled with precision. Manifolds, piping and cooling distribution units (CDUs) work together to ensure that the immense energy drawn by GPUs and CPUs is matched by equally robust cooling capacity. By demonstrating these systems in practice, the industry can demystify AI infrastructure and make it tangible rather than abstract.
For Africa, investing in resilient infrastructure is a strategic move. Data centres embracing scalable solutions can host advanced AI systems, attract investment, foster local innovation and support various industries like healthcare and energy. Futureproofing today is essential for African data centres to seize AI opportunities and lead in the global digital economy.
