Is AI Green Or Good For Green?

AI is the new fire, promising to forge a future full of endless possibilities. The use cases for AI to build and sustain utopia are plentiful. We’re told it will model our climate with perfect clarity, optimise our power grids, and design miraculous new materials. But what fuels this fire? It is fuelled by a hunger for energy (who guessed data? Tired of saying data is the new oil? Me too). Is it starting to burn down the forests we’re trying to save and do harm to the environment faster than it is doing good?

This is the central paradox of our age: the tool for our salvation is powered by our original sin of consumption. The explosive growth of AI has ignited a kilowatt arms race, a frantic scramble for power that is reshaping global markets and corporate strategy. The titans of technology, once content to be mere customers of the grid, are now becoming energy companies, buying and building power plants to feed their machines. This isn’t just about keeping the lights on.

The seductive vision of a green data frontier in Africa is emerging as a solution, a continent of untapped sun and geothermal heat ready to power our computational dreams. But is this a vision of the future or a mirage in the desert? Are we blinded by hope and flashy news?

Whose Got The Power?

It is estimated that power consumption in data centres in the US will triple by 2028; that’s not too far away. Will Africa be the solution in 3 years? Are we layering a hyper-advanced, power-hungry industry onto nations struggling with broken grids, deep-seated energy poverty, and geopolitical shifts?

The hard truth is that technology cannot fix a broken culture of governance. The path forward isn’t just about finding more energy. We can easily find a lot of energy in Africa, and we can build power plants to extract, convert and distribute all this power, but that doesn’t mean the large data centres with all the AI chips serving a growing market of billions of people will move into Africa just to save on power alone; it’s a lot more complicated than that.

The scale of AI’s thirst is staggering. A single ChatGPT query can use ten times the electricity of a simple Google search. Training one major AI model can consume as much energy as 130 U.S. homes do in a year and use over 700,000 litres of fresh water for cooling.

Faced with this reality and the creaking inadequacy of public grids, Big Tech has made a stunning pivot. They are no longer just buying power; they are buying power plants. Microsoft is helping restart a dormant reactor at Three Mile Island, the site of America’s most infamous nuclear accident. Amazon is spending $650 million to acquire a data centre campus powered directly by a nuclear plant.

It’s about speed. The public grid, with its decade-long queues for connection and permitting, is a bottleneck in a race measured in months, not years anymore. By building their own “behind-the-meter” power, these companies can bypass the system entirely, creating private energy islands to fuel their ambitions. But this creates a dangerous precedent: a two-tiered energy future where the wealthiest corporations enjoy clean, reliable power, while the public is left with a strained, more expensive grid, potentially subsidising the AI boom with their own utility bills.

Renewing Energy Is A Superpower 

Africa is a renewable energy superpower, holding 60 per cent of the world’s best solar resources and vast, untapped geothermal potential. The vision is intoxicating, a continent leapfrogging fossil fuels to power the future of AI with clean energy.

If Big Tech sets up data centres with their own energy grids, bypassing the existing African grid,s they would be able to meet their demand needs, but what does that leave the continent with? Broken plumbing that won’t be improved for a long time, as there is less incentive to do so.

The brute-force approach to bigger models, more power, is a dead end. The most meaningful solution to AI’s environmental crisis won’t come from an endless race for more power plants, but from a fundamental shift in how we design intelligence itself. Africa can have these Big Tech giants set up data centres, but instead of allowing them to build their own private energy grid, why not have them support, improving the current grid to make the energy more reliable, more carbon neutral, more accessible, cheaper and just more energy for everyone.

Maybe Africa itself tackles this problem from a model perspective, instead of trying to build bigger and bigger, as the global south Africa can work on building smaller, more intelligent models that require less power and therefore become sustainable for the power grid. This could be a blueprint that the whole world could even adopt from us.

This is the real AI revolution, not just building a machine that can think, but building a system that thinks sustainably. The choice is ours. We can continue down the path of mindless consumption, fuelling a machine until it burns through our resources. Or we can become architects, consciously designing a new, more intelligent foundation for the future we claim to want.

In The Hitchhiker’s Guide To The Galaxy, a supercomputer was built to answer the ultimate question of life, the universe and everything. The answer was 42. The answer was unsatisfying because the actual question itself was unknown, and so an even bigger computer was built to determine the question. Let us not turn this error in fiction to an error in real life.