Africa Must Fix Power to Compete in the AI Data Centre Race

Africa’s AI future will not be decided only by startups, talent, regulation, or whether young people can build the next great app. It may be decided by something far more basic: electricity.

That reality became impossible to ignore after reports that Microsoft’s planned $1 billion AI data centre project in Kenya, developed with UAE-based G42, had run into a major power problem. The project was originally announced in 2024 as a landmark East African cloud and AI infrastructure investment, with plans for a geothermal-powered facility that would support Microsoft Azure services in the region. But Kenyan President William Ruto has since warned that the proposed facility, at its larger scale, would require so much electricity that the country would effectively have to “switch off power for half the country” to run it. Kenya’s national installed capacity is reportedly around 3,000 to 3,200 megawatts, while peak demand is already above 2,400 megawatts. A one-gigawatt AI data centre would therefore consume a huge share of the country’s available power. 

That single statement captures Africa’s AI infrastructure dilemma better than any policy speech could.

The continent wants to be part of the AI economy. It has the users, the developers, the mobile adoption, the fintech ecosystems, the data-hungry businesses, and some of the fastest-growing digital markets in the world. But AI does not run on ambition. It runs on compute. Compute runs on data centres. And data centres run on enormous, stable, affordable electricity.

That is where Africa’s dream begins to collide with its grid.

The Kenya project is especially revealing because it was supposed to be one of the cleaner, smarter versions of AI infrastructure. It was tied to geothermal power in the Olkaria region, one of Kenya’s strongest renewable energy assets. This was not a reckless diesel-powered server farm dumped into a fragile grid. It was meant to be an example of how Africa could use renewable energy to leapfrog into the AI era. Yet even that model has struggled because the scale of modern AI infrastructure is now so large that even relatively power-rich African markets can be overwhelmed. 

This is the hard truth: the global AI race has become an energy race.

The International Energy Agency projects that global data centre electricity consumption will roughly double by 2030, reaching around 945 to 950 terawatt-hours, with AI-focused data centres growing even faster than the wider sector. Goldman Sachs Research projects data centre power demand could rise by about 175% by 2030 compared with 2023 levels effectively adding the equivalent of another top-10 power-consuming country to the global electricity system.

That is why the world’s biggest technology companies are no longer just buying chips. They are chasing electricity, land, cooling, transmission lines, water rights, nuclear power deals, renewable energy contracts, and battery storage. In the United States, battery storage companies are already seeing rising demand from AI data centres, partly because grid connection timelines in some regions can take three to seven years, far longer than the 18 to 24 months it may take to build a data centre.

This is happening in rich countries with far more mature grids than most African economies. So the challenge for Africa is not theoretical. It is structural.

South Africa remains the continent’s strongest example of what cloud infrastructure can look like when power, connectivity, enterprise demand, and market size align more convincingly. Microsoft opened Azure cloud regions in Johannesburg and Cape Town in 2019, Oracle opened its first African cloud region in Johannesburg in 2022, and Google Cloud opened its Johannesburg region in 2024. AWS also has an Africa region in Cape Town. In other words, when hyperscalers come to Africa, they overwhelmingly come through South Africa first. 

But even South Africa’s data centre expansion is not free from infrastructure concerns. Equinix’s proposed Cape Town data centre development is facing opposition over environmental issues, including questions around electricity demand, water use, emissions, and backup power systems. The planned facilities could reportedly require up to 160 megawatts of power combined, a major load in a city that has already experienced serious water scarcity. 

That should tell African policymakers something important: data centres are not ordinary office buildings. In the AI age, they are industrial-scale energy assets.

Nigeria is perhaps the clearest example of Africa’s opportunity and its contradiction.

With a population now well above 230 million and estimated by Worldometer, based on UN data, at about 242 million in 2026, Nigeria should be one of the most attractive AI infrastructure markets in the world. It has one of Africa’s largest internet populations, a young demographic base, a booming fintech sector, massive mobile usage, a growing startup ecosystem, and rising demand for cloud, streaming, payments, government digitisation, and AI tools. DataReportal estimates Nigeria had about 109 million internet users at the end of 2025, while other telecom-linked estimates put active internet subscriptions significantly higher depending on methodology. 

On demand alone, Nigeria should be a natural home for AI data centres.

But power changes the equation.

Nigeria has installed generation capacity above 13,000 megawatts, but effective available generation is often far lower. Recent reports put available generation around 4,000 to 5,000 megawatts, with Reuters reporting in February 2026 that gas supply shortages had reduced national power generation to about 4,300 megawatts, prompting load shedding and reduced distribution. Another 2026 power-sector analysis put available capacity at roughly 4,901 megawatts in January, only about 36% of installed capacity. 

That means Africa’s largest population may be trying to compete in the AI age with less grid power than a cluster of large data centres in Virginia, Texas, or Ireland might soon require.

This is the uncomfortable comparison Nigerian leaders need to sit with. A country of more than 230 million people, with tens of millions of internet users and some of Africa’s most active digital consumers, should not be debating whether it can reliably generate 5,000 megawatts. It should be planning industrial-scale clean power corridors for cloud, AI, manufacturing, transport, and cities.

Nigeria already has data centre players such as MainOne, Rack Centre, and other local operators, and reports show growing interest in expanding domestic capacity due to internet growth, 5G rollout, cloud adoption, and data localisation demands. But the market still faces the same core constraint; reliable power is expensive, and many operators must build around the grid using dedicated energy arrangements, backup systems, diesel, gas, or hybrid power. 

That makes local data hosting more expensive, which makes hyperscalers cautious, which keeps more African workloads outside the continent, which then weakens the business case for even more local infrastructure. It becomes a cycle.

This is one reason big cloud companies have not built data centres across Africa at the same pace they have in Europe, North America, and parts of Asia. The issue is not just demand. It is the combined economics of power availability, grid stability, fibre density, enterprise spending, regulatory clarity, currency risk, security, cooling, land, and long-term anchor customers. A hyperscale cloud region is a long-term commitment. Companies do not put billions of dollars into a market simply because the population is large. They need confidence that power will be available every hour of every day, at a predictable cost, with enough redundancy to protect uptime.

That is what Africa must understand. AI infrastructure does not tolerate “almost stable” power.

A normal office can survive outages with inverters, diesel generators, and patience. A hyperscale AI data centre cannot operate on excuses. Training and inference workloads require stable, high-density power. Cooling systems must run continuously. Network equipment must stay live. Backup systems must be ready. If power is unreliable, the cost of redundancy rises. If the cost rises too much, the region becomes unattractive compared with places that can deliver cheaper, cleaner, more reliable electricity at scale.

This is why Microsoft’s Kenya challenge matters far beyond Kenya.

It is a warning to the entire continent.

Africa has spent years talking about data sovereignty, local cloud regions, digital transformation, AI policy, and the need to stop exporting African data to foreign servers. All of that is valid. But sovereignty without electricity is rhetoric. You cannot regulate your way into the AI infrastructure race if your grid cannot power the machines.

The opportunity is still enormous. Africa has advantages that matter: abundant solar potential, geothermal resources in East Africa, hydro resources in several regions, gas reserves in countries like Nigeria and Mozambique, large young populations, growing digital demand, and strategic geography between Europe, the Middle East, and the Americas. The continent also has a chance to build cleaner AI infrastructure from the start instead of copying the fossil-heavy path of older industrial economies.

But that requires a different kind of policy thinking.

African governments should stop treating power, cloud, fibre, and AI as separate sectors. They are now the same strategy. A national AI plan that does not include a serious energy plan is incomplete. A data centre policy that does not include transmission upgrades, renewable generation, gas reliability, grid reform, and industrial tariffs is just paperwork. A digital economy agenda that ignores electricity is a slogan.

For Nigeria, the path should be clear. The country should identify dedicated AI and cloud infrastructure zones near reliable power sources, gas corridors, renewable clusters, and submarine cable landing points. Lagos has connectivity advantages, but power and congestion issues remain real. Other regions could become compute hubs if planned properly around energy-first infrastructure. Nigeria should not wait for hyperscalers to solve this. It should create bankable energy-backed digital infrastructure zones where local data centre companies, telecom operators, banks, cloud providers, government workloads, and AI startups can cluster around guaranteed power.

Kenya’s situation also offers a lesson. Its geothermal advantage is real, but even renewable power must be scaled with industrial demand in mind. A country cannot simply announce a gigawatt-scale AI facility and hope the existing grid absorbs it. It needs transmission planning, phased capacity, local load balancing, and a realistic view of what citizens and industries already need.

South Africa’s lesson is different. It shows that cloud regions can come to Africa when market fundamentals are strong enough, but it also shows that data centres will increasingly face scrutiny over power and water. That means future African AI data centres must be built with transparency, sustainability, and local benefit in mind. Communities will not accept massive electricity and water users if they feel the benefits flow only to multinational platforms.

And that brings us to the deeper point.

Africa should not want AI data centres simply for prestige. It should want them because local compute changes the economics of innovation. It reduces latency. It improves data residency. It supports local AI model development. It helps banks, hospitals, schools, startups, governments, and media companies run digital services closer to users. It creates room for African languages, African datasets, African cybersecurity systems, and African AI applications to be built with less dependence on infrastructure thousands of miles away.

But if Africa gets the power question wrong, it will become a consumer of AI rather than a producer of AI.

That is the real danger.

The AI race is not just about who uses ChatGPT, Gemini, Claude, or whatever comes next. It is about who owns the compute layer underneath those tools. Countries with power will host the models, train the systems, store the data, attract the cloud companies, and build the next generation of AI businesses. Countries without power will remain users on someone else’s platform.

This is why Microsoft’s stalled Kenya data centre should not be read as a failure. It should be read as a wake-up call.

Africa does not lack intelligence. It does not lack ambition. It does not lack young people ready to build. What it lacks, too often, is the physical infrastructure that turns digital ambition into industrial power.

In the AI age, electricity is no longer just a development issue.

It is national competitiveness.

And for Africa, getting power right may be the difference between joining the AI race and watching it happen from the sidelines.