By Eamonn Ryan, based on a presentation by Wojtek Piorko at the seminar AI Unlocked: What ‘AI-ready’ means for 2026 decisions
Much of the conversation around artificial intelligence (AI) centres on algorithms, applications and automation. In Africa, however, the more immediate question is energy: the power required to run the data centres that underpin it all. Behind every AI model sits a physical reality of electricity supply, cooling demand and infrastructure resilience – the true foundation of the digital economy.
As AI systems become more demanding, the design and operation of this infrastructure – particularly in Africa – must evolve rapidly to keep pace with global trends.
AI Unlocked: What ‘AI-ready’ means for 2026 decisions. © RACA Journal
The rise of high-density computing
The growth of GPU-based infrastructure is dramatically increasing power density within data centres. A traditional server might consume around 300 watts. Modern GPU-enabled servers can require four times that amount or more, and the trend is accelerating. Historically, racks in many data centres operated at around 13kW. In Africa, the average is often closer to 5kW per rack. But AI workloads are changing those numbers rapidly.
Next-generation GPU platforms are pushing rack densities far beyond previous limits, potentially reaching hundreds of kilowatts per rack in the coming years.
This shift has several implications:
- Cooling systems must remove far greater amounts of heat
- Power delivery systems must handle dramatically higher loads
- Structural considerations, including floor loading, become more complex
In short, the architecture of the data centre itself is evolving.
Power and cooling: Africa’s critical constraints
One of the biggest questions surrounding AI deployment in Africa is power availability. The continent’s total power generation capacity has roughly doubled over the past decade and is projected to reach around 260 gigawatts. Yet generation capacity alone does not solve the problem.
Distribution remains a major challenge. Many regions have abundant natural resources, including significant hydropower potential, but infrastructure limitations prevent that energy from reaching where it is needed.
For data centre operators, this creates both risk and opportunity. High-density computing also requires advanced cooling solutions. As AI workloads grow, liquid cooling is increasingly becoming part of the design strategy for new facilities.
This means the infrastructure inside the data centre – especially above the racks – is becoming far more complex. Traditionally, the space above server racks contained little more than cable trays. Today, it may host fibre networks, cooling systems, fluid distribution lines, containment systems and a range of other critical components.
Modular infrastructure for a fast-moving industry
Another key trend in modern data centre design is modularity. Instead of constructing facilities entirely on-site, many components can now be pre-engineered, manufactured in factories and delivered ready for installation.
This approach offers several advantages:
- Faster deployment, allowing infrastructure to scale quickly as demand grows
- Improved quality control, with factory testing before delivery
- Reduced construction complexity and waste on site
- Greater scalability, allowing operators to expand incrementally
Modular systems can apply not only to buildings themselves but also to internal infrastructure – power systems, cooling systems and even rack-level technologies. In an industry where technology evolves rapidly and demand patterns shift quickly, this flexibility is becoming increasingly valuable.
Planning for an uncertain future
AI infrastructure development is moving at extraordinary speed. Chip manufacturers are releasing new generations of processors every year, dramatically increasing compute capacity and power requirements.
This rapid innovation raises new questions about sustainability and lifecycle management. What happens to older hardware as new generations arrive? How can infrastructure remain adaptable as technologies evolve? And how can operators ensure that investments remain viable in a rapidly changing landscape?
These are questions the industry is only beginning to answer. What is clear, however, is that the future of AI infrastructure will be shaped by a combination of higher densities, smarter energy management and modular deployment models.
A critical moment for Africa’s digital future
Africa currently operates only a fraction of the global data centre capacity required to support its digital growth. Yet the demand for digital services, and increasingly for AI-enabled services, is rising inexorably.
Expanding infrastructure will require investment, regulatory alignment, improved power systems, and the development of local technical skills. But it will also create significant opportunities. The next generation of data centres will not simply support digital services. They will underpin the continent’s ability to participate fully in the global AI economy.
And as the pace of technological change accelerates, the ability to deploy scalable, efficient infrastructure may become one of the most important competitive advantages of all.
