The artificial-intelligence boom is producing a problem that has little to do with algorithms: it needs electricity, and a lot of it. The rapid construction of data centres to power AI systems is placing growing pressure on electricity grids, water resources and local infrastructure worldwide.
For years, public debate around AI focused on jobs, misinformation, privacy and the future of work. Now a more basic question is becoming harder to ignore: where does the physical power to run AI come from? Large data centres draw electricity continuously and generate enormous heat, requiring heavy cooling systems. Communities hosting new facilities are increasingly questioning whether the economic benefits justify the strain on power, water, roads and other infrastructure.
The scale is compounding fast: more computing capacity means more servers, more servers mean more electricity, and more electricity means new generation and transmission infrastructure — often at significant cost. That raises a political problem: if private technology companies build the facilities but public utilities have to upgrade grids, ordinary electricity customers and taxpayers may end up bearing part of the bill.
Large US, European and Gulf-based tech firms have all announced multi-billion-dollar data-centre build-outs through 2026, several sited specifically near cheap power or water sources — a pattern already drawing local pushback in parts of the US, Ireland and the Gulf over rising electricity costs and water use.