AI’s next crisis isn’t electricity or even sentience, it’s something far more essential to both humans and machines
The most powerful billionaires in AI are staring down the barrel of an existential crisis as they scramble to keep the AI boom growing exponentially.
In Frank Herbert’s best-selling novel Dune, a peculiar and dystopian premise was put to the world.
What if water became so scarce that it turned into a currency, culture, and religion? A planet where the super-rich schemed to control the most crucial finite resource known to man.
Similar themes were touched on in Mad Max, and we all know how grim that depiction of Australia looked.
Unfortunately, art really does imitate life in some cases.
About 25 per cent of countries on Earth face chronic water scarcity – and that’s just drinking water.
But now, the world’s most influential tech figures are in a bind over exactly how to source enough just to keep their AI servers afloat.
Nobody expected that we’d be having this conversation a few short years into the AI boom, but Europe is steaming towards an ugly crossroad.
For all the latest science and technology news — download the news.com.au app direct to your phone.
MORE: How AI is helping Aussies find work
With record heatwaves, raging wildfires, and prolonged droughts sweeping across the continent, the crisis no longer lies in the skies or on the ground. The problem is in the digital cloud, and specifically, how to keep that cloud from overheating.
While millions around the world sit glued to their screens incessantly using artificial intelligence for everyday tasks, AI data centres have developed their own enormous addictions.
According to experts from leading data and analytics firm GlobalData, the issue of water scarcity has now outstripped skyrocketing electricity demands.
“Climate change has produced weather patterns from wildfires to flash floods that are becoming more extreme worldwide,” Robert Pritchard, Principal Analyst at GlobalData, says.
“Superheated ground results in less rain getting absorbed into aquifers and instead running over dry soil, often destroying lives and livelihoods.
“Climate change also means that the requirement for water to cool the ever-expanding base of data centres is likely to become an issue that needs addressing.”
It gets a bit wild once you start crunching the numbers.
A 2023 OECD study estimated that AI systems will consume between 4.2 and 6.6 billion cubic metres of water per year by 2027. That’s more than the annual water use of Denmark, or nearly half the entire United Kingdom.
Most of it is used to cool high-performance computing clusters powering large language models. But as models grow larger and demand scale ups, so does the drain on freshwater supplies.
With tensions mounting, some cities have actually started pushing back. Singapore and Dublin have already paused or blocked new data centre construction over fears that water and energy systems were becoming dangerously strained.
“Data centre providers are using technology innovations to try to address the energy demands of their growth, particularly given the explosion of artificial intelligence,” Mr Pritchard said.
While alternatives like air cooling or immersion systems exist, water remains the industry’s most cost-effective and reliable solution.
“As with most things in tech, solutions will be found for the emerging challenges. The problem, however, is that any solutions must be set within the context of climate change and the immediate impact it is having on citizens’ lives,” Mr Pritchard continued.
“It is a political issue and a social issue, not just a technology issue.”
I know what you’re thinking. Why not seawater?
Unfortunately, the issue isn’t as simple as scooping out a few litres from the Pacific every year.
Seawater just isn’t practical for cooling data centres because it’s highly corrosive and requires extensive treatment before it can be used safely.
Salt content can damage pipes, heat exchangers, and other infrastructure, leading to higher maintenance costs and reliability risks. Even when used indirectly, it still needs to be filtered, desalinated, or chemically treated to avoid corrosion, scaling, and biofouling.
In short, the process of treating seawater for cooling use is just too costly. And unlike electricity, which can at least be temporarily backed up with generators, freshwater has no substitute.
“There is no water equivalent of carbon credits to hide behind, obfuscation tactics that have often saved blushes among greenwashers,” Mr Pritchard said.
“On top of that is the potential threat of terrorist attacks on critical water infrastructure, which looks like far more of an important issue.”
The problem is humanity’s crippling addiction to technological progress and automation isn’t going to stop any time soon. The data centres will continue to be built and more GPT updates will be coming in hot. Too hot, evidently.
Welcome to Arrakis, folks.
Originally published as AI’s next crisis isn’t electricity or even sentience, it’s something far more essential to both humans and machines
