Wishes And Dreams Will Not Move Artificial Intelligence Forward In Africa!

If wishes were horses, the sages say, beggars will ride. That is exactly what African countries are doing when it comes to Artificial Intelligence (AI). Unfortunately, policies do not get implemented when you sit down to dream or wish them to happen.

Like everything that Africa has failed to do in the past, it does look like, governments across the continent are not looking at what can make AI become part of what will drive their economies. While other countries are taking steps to make the implementation of AI policies viable, African governments are just talking about where AI will take their countries. Issues that will hinder its implementation are not being considered. One area that has not been talked about all these years is whether or not Africa can cope with the energy requirements and water supplies needed for AI data centres.

Alex Chenglin Wu, founder and chief executive officer (CEO) of DeepWisdom, says AI data centres require reliable electricity supply and can draw enormous amounts of power, pointing out that, “a medium-to-large facility, for example, can use almost as much electricity as a small city.”

Wu said in countries with more developed infrastructure—such as South Africa, Kenya, Morocco, and Egypt—powering a limited number of AI data centres is entirely realistic, adding that, “if these facilities scale across the continent, they could place additional pressure on residential electricity supplies. This means AI data centre expansion cannot rely solely on existing public grids.”

“With a solid engineering plan, it is entirely possible to deliver stable power to a small number of large AI data centres. Such a plan would combine dedicated power generation, energy storage systems, and advanced cooling technologies,” he said.

This means that Africa governments should not just be talking about the importance of AI, and how to implement it. They should also be looking at how to provide the energy required. In addition, on a continent where water is scarce, it also means that, the provision of water for use at the data centres should also become an issue that should be discussed now.

Wu says if more data centres are built, Africa’s abundant renewable energy resources will become crucial, adding that, “in regions with strong solar and wind potential—such as South Africa, Kenya, and Morocco—clean power plants could be developed near data centres, including solar farms, wind farms, or small gas-fired power plants.”

He said electricity would be supplied first to the data centre, with any surplus fed back into the grid, explaining that “these power plants could also be supported by microgrids and battery storage.”

“Energy can be stored during the day and released at night. If the public grid fails unexpectedly, batteries can provide backup power for a period, helping ensure that AI servers remain online,” he added.

Touching on the water issue, Wu said cooling is another major challenge, because Data centres require substantial cooling capacity, and water-based systems can consume large volumes of water. For this reason, he suggested more efficient, water-saving methods—such as liquid cooling, advanced air-conditioning systems, and integrated thermal management—should therefore be prioritized.

AI infrastructure also carries a significant water cost. Servers generate substantial heat, and evaporative cooling towers remain the most widely used method of thermal management in data centres. A single training run for GPT-3 at Microsoft’s U.S. data centres consumed approximately 700,000 litres of freshwater. In 2024, Google’s total water consumption across its data centres and offices reached approximately 8.1 billion gallons.

“Many of the energy challenges Africa may face in building data centres are, at their core, engineering problems. With careful planning and the right investments, they can be addressed,” he added.

In a commentary for the Brooking Institute, Chinasa Okolo, founder and scientific director Technecultura, said after the “Global AI Summit On Africa,” in Rwanda last year that, there are several structural challenges that threaten the harmonization of Africa’s AI ambitions.

Okolo said conflicts, such as violent extremism in the Western Sahel, the departure of authoritarian-ruled countries from the ECOWAS regional bloc, and the humanitarian crisis in Sudan, continue to impact regional harmony.

“These divisions raise concerns that efforts to position the continent as a “major AI player” may inadvertently lead to further stratification,” she said, adding that, “such fragmentation could advance the goals of larger countries with more economic resources to devote to AI, while marginalizing smaller countries with fewer resources, a pattern that mirrors existing economic inequalities within the continent.”

Okolo said a growing number of data centres (3,069 data centres already operate in the U.S., with an additional 1,489 planned or under construction) are “hyperscale” data centres. —Driven by the AI boom, these sites can sprawl across thousands of acres and consume vast amounts of power.

Currently, there are about 68 hyperscale facilities that each use at least 50 MW of electricity—as much as a small city. The biggest can use 5 GW or more. —Hundreds more hyperscale facilities may soon be built across the U.S. with 267 planned data centres, each using 50 MW or more.

Hyperscalers are massive cloud services providers that build and manage extreme-scale data centres to deliver on-demand computing, networking, and storage.

Wu agreed with Okolo and said, in terms of power use, a typical data centre runs at about 10–25 megawatts (MW), while a hyperscale, AI-focused facility can exceed 100 MW, using as much electricity each year as roughly 100,000 homes.

“As these facilities expand, so does their reliance on fossil fuels, which currently supply nearly 60% of global data-centre electricity. Rising energy demand is expected to drive a 7.3% increase in natural gas output between 2025 and 2027,” he added.

Wu said this means the rapid growth of AI infrastructure may prolong dependence on conventional energy sources, as the expansion of green energy capacity has so far been unable to keep pace with surging demand.

He said another issue that must be looked at is the fact that, AI training depends on massive volumes of digital information that must be stored, processed, and computed in data centres. Globally, hyperscalers are projected to invest up to $7 trillion in data centre infrastructure through 2030.

There is also the need to look at the finance side of AI operations. He said after hitting record levels in 2025, capital expenditure on AI infrastructure is expected to rise by more than 60% in 2026. According to CNBC and Bloomberg, Alphabet, Microsoft, Meta, and Amazon are expected to spend nearly $700 billion on AI‑related capital expenditure, much of it directed toward data centres, specialized chips, and networking infrastructure.

“Of that, Amazon is projected to spend about $200 billion, Microsoft around $145 billion, and Meta between $115 billion and $135 billion, while Alphabet is expected to invest roughly $175 billion to $185 billion,” Wu added.