Everyone thinks the AI boom is about chips. Faster chips. Bigger chips. More chips. But that’s already yesterday’s problem.
The real constraint is power (electricity) – and that quiet shift is about to reshape energy markets, revive nuclear power, and open a rare opportunity for everyday investors.
We’re walking into a new kind of power crisis
Not one triggered by oil embargoes or hurricanes, but by artificial intelligence quietly transforming data centres into 24/7 industrial-scale power consumers.
AI models don’t sleep. They don’t idle. Training and running large models require thousands of high-performance chips operating continuously, drawing enormous amounts of electricity and generating heat that must be cooled around the clock. As AI scales from experiments to core infrastructure, its energy footprint explodes.
Nvidia CEO Jensen Huang recently put it bluntly: energy – not chips – will soon be the main constraint on AI growth. His warning wasn’t theoretical. He’s seeing it firsthand as customers struggle less with accessing GPUs and more with securing enough reliable power to turn them on. Huang has suggested that data-centre operators will increasingly turn to small nuclear reactors within the next few years.
That may sound extreme…but the data says otherwise
Global data-centre electricity consumption is on track to more than double by 2030, reaching roughly 945 terawatt-hours per year – about the same as Japan’s total electricity usage.
The International Energy Agency (IEA) has been clear…
AI is the single most important driver behind this surge. Power demand from data centres is growing at roughly 15% per year, around four times faster than overall global electricity demand.
In certain regions, the strain is already visible. Northern Virginia, the world’s largest data-centre hub, now has server farms consuming over 25% of local electricity capacity. Utilities there are warning that grid expansion is struggling to keep pace. Infrastructure designed to serve homes, offices, and shopping centres was never meant to support dense clusters of energy-intensive computing facilities.
This brings Big Tech to an uncomfortable conclusion…
Renewables alone can’t carry the power load…
Wind and solar are clean and increasingly cheap, but they’re intermittent by nature. AI workloads – from cloud computing and financial trading to defence and healthcare systems – require power every second of every day.
Batteries help smooth short-term gaps, but current battery technology cannot provide multi-day or industrial-scale back up at reasonable cost.
Building new gas-fired power plants is faster, but it comes with regulatory delays, public opposition, and rising emissions.
New transmission lines face similar challenges and often take a decade or more to permit and construct. If AI’s power demand is met primarily with fossil fuels, emissions rise sharply.
One estimate suggests that if 60% of incremental AI electricity comes from fossil sources, global CO₂ emissions would increase meaningfully from data centres alone – an outcome that is politically and reputationally unacceptable for governments and hyperscalers.
That’s why nuclear power is back in the picture!
Nuclear power offers something no other energy source can: reliable, 24/7, carbon-free baseload power at massive scale. And advanced designs like small modular reactors and microreactors change the economics.
These systems can be built faster, deployed in smaller increments, and located closer to where the power is needed – including near data-centre campuses.
The shift is already underway. Amazon Web Services (AWS) has begun backing nuclear projects to secure long-term power for its cloud infrastructure. Governments are extending the lives of existing reactors, streamlining approvals, and reopening discussions around new builds.
For investors, the implications are clear…
AI doesn’t just need software and chips. It needs physical energy inputs – uranium, fuel conversion, enrichment, reactor components, and life-extension services for existing plants.
This widening gap between AI’s power needs and grid reality is “The Crunch.” Nuclear is the only scalable, always-on solution capable of closing it. And that convergence may turn out to be one of the most important and underappreciated investment themes of the next decade.
And interestingly, two very different kinds of companies sit right at the centre of this convergence.
One doesn’t sell electricity at all. It builds and fuels the nuclear machines themselves, supplying mission-critical systems to governments, space programs, and next-generation reactors.
The other already owns and operates the largest fleet of nuclear plants in the country and is quietly signing direct deals with Big Tech to keep AI data centres running without interruption.
Different roles. Same profit-opportunity.
Simply put – as the AI energy crunch accelerates, these “picks-and-shovels” players in the nuclear revival may prove far more important and far more valuable than most investors currently realise. Follow South African Investor for more on the future of energy and AI.
Not a subscriber to Money Morning?
You can get free daily recommendations like these with Money Morning eletter. Just sign up here.
