Datacentres are a ticking time bomb. We must make sure AI’s benefits outweigh the costs | Nicki Hutley
TL;DR
Nicki Hutley ties the AI boom to datacentre expansion: the facilities use large amounts of power and water, emit waste heat and add pressure to climate and infrastructure policy. The article says more than 10,000 datacentres are active worldwide. That number could grow 3.5 times, with estimated investment of US$7tn. Australia is the case study: 286 active or planned centres, Anthropic interest and forecasts that datacentres could triple national power and water consumption by 2030.
Nauti's Take
The piece hits a sore point: too many AI debates treat datacentres like neutral internet plumbing. They are not.
If a site strains local grids, uses scarce water and creates few long-term jobs, calling it infrastructure is not enough. A cleaner test would ask which AI use case justifies which physical footprint.
Travel planners and meme generators look very different under that lens than medical imaging or grid optimisation.
Briefingshow
The debate pulls AI out of the productivity slide deck and into the world of constrained grids, water budgets and permits. For companies, scaling AI is not just a model-access decision; it also imports compute demand, energy pressure and political friction. The piece is opinionated, but the core question lands: who benefits, who pays and who checks before the buildout happens?
Sources