Google's electricity use rose 37% in 2025, its biggest ever jump, driven by an AI buildout outrunning grid decarbonisation. The AI tools marketers use feel free but carry a real and growing cost. Treat AI as a cost to manage, not a free tap left running.
Every AI feature you switch on has a power bill attached. Someone is paying it, and increasingly it is the planet.
Google's electricity use rose 37% in 2025, the largest single year jump in the company's history. Its 2026 environmental report, released on 30 June, shows total power consumption is now more than 250% higher than in 2019. The driver is the AI buildout. Google said plainly that its AI infrastructure is expanding faster than the grid is decarbonising.
The emissions picture followed. Supply chain emissions grew 25% year on year, with data centre construction alone adding around 2.3 million tonnes of carbon dioxide equivalent, much of it from semiconductor suppliers on carbon heavy grids. Google still matched 100% of its electricity with renewable purchases for a ninth straight year and signed deals for more than 12 gigawatts of new clean energy, but the demand curve is outrunning the supply of clean power.
This is the cost side of the AI story that marketing rarely talks about. The tools are cheap to use and expensive to run. The gap between those two facts is being absorbed, for now, by the companies building the infrastructure.
Why it matters
Marketers are pouring AI into everything, content, targeting, reporting, creative. Most of it feels free at the point of use. It is not. The energy cost is real, it is growing, and it is starting to attract regulatory and reputational attention. For any brand with a sustainability position, unbounded AI use is a claim risk waiting to happen.
The rise in Google's electricity use during 2025, its biggest annual jump ever, driven by the AI buildout. Source: Google 2026 Environmental Report.
There is also a commercial read. Compute this expensive does not stay free forever. The cost will find its way into pricing eventually, so building your whole operation on cheap AI is a bet on subsidies lasting.
What to do about it
The buildout is not slowing. Neither is the power draw. The marketers who treat AI as a cost to manage, not a free tap to leave running, will be the ones still standing when the bill lands.