AI's Growing Appetite for Energy and Water — Why It Belongs on Every IT Leader's Radar

2026 coverage is putting numbers on something the industry has hinted at for a couple of years: demand for electricity to run AI is growing far faster than infrastructure is greening. Training a single large model can use as much power as around a hundred U.S. households consume in a year — and a less visible but equally real resource, fresh water for data-center cooling, is climbing right alongside it.

Major cloud providers are responding — hiring hundreds of energy specialists, shifting from evaporative to direct-to-chip and immersion liquid cooling, and getting regulatory approval to trade electricity directly. For a small or mid-sized business that simply plugs into someone else's cloud, though, none of that is visible unless you ask your provider directly.

It's another argument for why an IT audit matters in the first place: without an explicit question about a provider's energy and water use, "default" infrastructure is rarely the environmentally sound choice. We build these questions into our COBIT audits — not because it's trendy, but because for a company with a climate mission, this is exactly the kind of detail that adds up.