When OpenAI’s Sam Altman recently suggested the AI industry might be in a bubble, he shocked everyone and no-one. The global stock market has nearly doubled in five years, driven almost entirely by 10 companies whose rise hinges on the promises of AI. While most understand that this isn’t sustainable, the transformative potential keeps investors betting big.
Beyond the headline-grabbing market valuations lies a more sobering reality about the astronomical resource consumption fuelling this AI arms race. Market capitalisation reflects speculative pricing, not actual value creation or destruction. Far better to track the tangible resources, such as chips, energy and human talent, that are being poured into this pursuit.
The numbers are staggering. Just four major tech companies — Meta, Google, Microsoft and Amazon — are projected to spend $324bn on AI infrastructure this year alone. That’s more than 1% of the US’s entire GDP now dedicated to building computer capacity for developing and operating new models.
Global AI infrastructure investment has reached about $2-trillion when including Chinese state enterprises ($500bn) and other global players. Yet despite this enormous capital deployment, practical benefits remain elusive. Massachusetts Institute of Technology research tracking 300 companies implementing enterprise AI found that 95% generated zero returns, despite an additional $30bn-$40bn investment.
The energy bill behind the AI dream
The energy costs tell an even more troubling story. Data centres consumed 1.5% of global electricity in 2024, projected to double by 2030. In the US, electricity prices have nearly doubled over three years, coinciding precisely with the AI development surge.
This energy crisis highlights the infrastructure vulnerability of the US. Whereas Chinese grids can absorb AI data centres’ energy demands, the US grid is ageing and under strain.
American Big Tech’s response has been to commission private power plants or lobby for enormous infrastructure upgrades. The University of Texas Energy Institute estimates grid modernisation could cost $5-trillion, money that might address more pressing social needs.
Lessons for SA
For SA grappling with its own energy challenges and considering AI regulation frameworks, these developments are instructive. Our ongoing load-shedding crisis and infrastructure constraints make US-style AI spending practically impossible and economically questionable.
Rather than chasing Silicon Valley’s capital-intensive approach, SA might benefit from focusing on practical AI applications that solve problems without requiring huge infrastructure investments.
The AI revolution may transform our world, but at what cost?
The global AI arms race’s resource demands suggest our regulatory discussions should prioritise efficiency and sustainability over raw computational power. Perhaps our infrastructure limitations could guide us towards more sensible AI development that delivers genuine value rather than speculative returns.
The AI revolution may indeed transform our world, but at what cost? As investors pour trillions into an uncertain future the rest of us bear the immediate burden through higher energy prices, strained infrastructure and opportunity costs that will shape our economy.
The question isn’t whether AI will change everything. It’s whether we can afford the bill when it arrives. Policy forethought must be at the centre of SA’s AI planning.
• Timcke is senior research associate: research at ICT Africa, research associate at the University of Johannesburg's Centre for Social Change, and an affiliate of the Centre for Information, Technology & Public Life at the University of North Carolina at Chapel Hill.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.