By Kevin Subramani, Director of Industry, Mission Critical Facilities & Aviation – Atana

Africa is entering the AI era with data centre build costs estimated to be 30 to 40% higher than in Europe or the United States. At the same time, infrastructure requirements are shifting faster than ever.

In this environment, the idea of “future-proofing” is no longer viable. The real priority is future-enabling – designing infrastructure that can adapt as technology, workloads and energy demands continue to evolve.

AI hardware is advancing at extraordinary speed. Rack densities are rising rapidly, cooling technologies are evolving, including the growing adoption of liquid cooling, and power requirements are scaling in ways that would have been difficult to predict even a few years ago.

Trying to design infrastructure today for a fixed, fully defined future risks getting it wrong — either by overbuilding expensive capacity too early, or by locking facilities into designs that cannot adapt as technology evolves.

Designing infrastructure that can evolve

Future-enabled infrastructure does not attempt to predict exactly what AI infrastructure will look like five or ten years from now. Instead, it focuses on building flexibility into the physical design of facilities, allowing them to adapt as technology changes.

Many of the most important decisions are made long before the first servers are installed. Design choices, such as higher ceiling heights, stronger floor loading, larger equipment lifts and service corridors or electrical systems designed to expand over time, can be made to appear minor during the design phase. However, in reality, they can determine whether a facility can support future generations of technology.

These design considerations can be relatively inexpensive to incorporate during the initial build, but extremely costly to retrofit later.

Importantly, these same design choices also have long-term sustainability implications. Retrofitting infrastructure to support new cooling systems or higher densities often carries a significant environmental cost, not just financial. Designing for adaptability from the outset reduces the need for future rebuilds, lowering both embodied carbon and operational waste over the lifecycle of the facility.

AI is reshaping infrastructure requirements

AI computing places very different demands on digital infrastructure compared with traditional cloud workloads.

“AI is driving completely new infrastructure requirements,” says Wojtek Piorko, Managing Director for Africa at Vertiv. “Power to a single rack can be up to ten times higher than traditional IT workloads, which means cooling, power distribution and overall facility design must be reconsidered. A system-level infrastructure approach needs to be applied.”

Much of today’s AI processing relies on graphics processing units (GPUs), which can perform many calculations simultaneously. While this capability enables powerful AI models, it also drives significantly higher power densities within data centres.

Higher-density computing environments generate far greater heat loads. This is accelerating the adoption of liquid cooling technologies capable of managing much higher thermal output than traditional air-cooled systems. Facilities that are designed with the ability to transition between different cooling approaches will be far better positioned to adapt as these technologies evolve.

These shifts are not only technical – they are deeply tied to sustainability. As AI infrastructure scales, so too does its environmental footprint. Power consumption, water usage for cooling, and carbon intensity are becoming critical design considerations, not afterthoughts.

In this context, future-enabled infrastructure is not just about flexibility – it is about building systems that can evolve towards greater efficiency and lower environmental impact over time.

Energy and connectivity shape success

AI workloads also introduce new dynamics for energy systems. Unlike traditional enterprise computing environments, where demand typically grows gradually, AI workloads can cause electricity consumption to ramp up extremely quickly.

This places additional pressure on power infrastructure – an important consideration in many African markets where energy supply and grid stability are already significant challenges. But the challenge is not only the availability of power; it is also the sustainability of that power.

In many markets, where grids can be both constrained and carbon-intensive, the question is no longer just how much power is available, but how responsibly it can be sourced and used. This is driving increased focus on renewable energy integration, energy efficiency, and smarter load management as part of data centre design.

But energy alone does not determine where successful digital infrastructure can be built. Connectivity ecosystems are equally critical.

“Data centres are nothing without connectivity,” says Marc Matthews, Director of Engineering and Head of PMO at Open Access Data Centres. “Large compute environments are important, but without access to global networks and subsea cables, the infrastructure cannot deliver real value.”

Facilities located near subsea cable landing stations and major connectivity corridors are therefore emerging as strategic digital infrastructure hubs.

A growing industry focus is also shifting towards a more lifecycle-driven approach – designing infrastructure not just for immediate performance, but for long-term efficiency and adaptability. This includes the ability to upgrade systems without major rebuilds, reduce stranded capacity, and improve performance as technologies evolve. In this sense, flexibility and sustainability are closely linked: infrastructure that can adapt is inherently less wasteful.

Africa’s AI opportunity

Today, an estimated 80 percent of Africa’s data is hosted outside the continent, highlighting both a major infrastructure gap and a significant opportunity to develop local digital ecosystems.

The scale of that gap becomes clearer when viewed globally.

“Global data centre capacity is now over 120 gigawatts,” says Piorko. “Africa accounts for well under 2 gigawatts – less than 1% of the global total – highlighting the scale of the gap with more mature markets.”

Africa’s opportunity in the AI ecosystem doesn’t lie in competing with the massive training clusters being built in the United States or China. Instead, the continent is well positioned to support AI inference – where trained models are applied to real-world tasks such as answering questions, generating images or analysing data.

These workloads increasingly need to run closer to users and data, making regional data centre infrastructure essential. Africa’s opportunity is not to replicate global hyperscale models, but to build infrastructure aligned with its own digital growth trajectory, while remaining part of the global AI ecosystem.

Building infrastructure for Africa’s digital future

Planning for the successful future of data centres in Africa also requires challenging one of the most common misconceptions: that simply building large data centres anywhere on the continent will automatically unlock opportunity.

Reliable energy systems, strong connectivity ecosystems, supportive regulatory environments and proximity to real digital demand will all shape where sustainable infrastructure growth occurs.

The AI era is changing how digital infrastructure must be designed and deployed. The data centres that succeed will not be the ones designed for a perfectly predicted future. They will be the ones designed to adapt as that future unfolds – efficiently, responsibly and sustainably.