Datacentres are a ticking time bomb. We must make sure AI’s benefits outweigh the costs | Nicki Hutley
TL;DR
Nicki Hutley warns in The Guardian that the datacentre boom is becoming a climate, cost and distribution problem: AI infrastructure consumes power, water, land and cooling while its public benefit often remains vague. The article says there are more than 10,000 active datacentres worldwide. That number could grow 3.5x, with estimated investment of 7 trillion US dollars. For Australia, Hutley points to 286 active or planned sites. Datacentre energy and water use there could triple by 2030.
Nauti's Take
The useful argument is not that AI datacentres are bad. It is that we too often treat them as neutral future machines while they pull very real resources from very real places.
AI can reduce congestion, improve medical imaging and help stabilize power grids. Another travel-planning chatbot does not justify a blank cheque for water use, fossil power and higher electricity bills.
Serious AI policy has to ask which compute serves public value and which compute mostly serves margin and marketing.
Briefingshow
The piece moves the AI debate from model hype to the physical bill underneath it. For users, the next AI wave depends not only on better tools, but on power grids, water rights, cooling, planning rules and prices. If governments wave projects through, households and regions may absorb the costs without automatically sharing in the productivity gains.
Sources