AI – how much energy will it really devour?

By Finance News On Mar 8, 2025

Culpeper County, Virginia, may be home to cows and rolling farmland but its future is looking more digital than rural.

Seven massive data centres are being built there, part of big tech’s relentless push into artificial intelligence (AI). And while AI promises innovation, it comes with a cost—one that residents, researchers and even energy analysts are struggling to quantify.

A new report from Nature reveals that while AI’s global energy footprint may not seem alarming, its local impact could be devastating.

Data centres are clustering in places like Virginia and Dublin, consuming as much power as entire cities.

The biggest challenge? Transparency. Researchers trying to calculate AI’s electricity use are left “grasping at straws,” says Alex de Vries, who has spent years analysing the digital economy’s environmental impact.

Even tech giants like Google and Microsoft, which claim to be working toward sustainability, are staying tight-lipped about their energy use.

AI’s hunger for power is undeniable. One study estimates that integrating ChatGPT-like AI into every Google search could require up to 29 terawatt hours annually – about 30 times the energy of a standard search.

Training large models like ChatGPT takes gigawatt hours. And as AI scales up, so do the power demands.

Some governments are stepping in. The EU now requires data centres to report their electricity use and Virginia has passed a transparency bill (though it stops short of mandating disclosure).

Meanwhile, energy grids in AI hotspots are already feeling the strain. Delays, price hikes and even revived nuclear projects are on the table to meet demand.

AI’s future remains uncertain. Will efficiency improvements curb its energy appetite, or will growing demand make it worse?

One thing is clear: the industry’s reluctance to share data is making it harder to predict—and even harder to control.

devour Energy