AI is eating the planet: Google's Gemini now burns more power than Denmark

The tech giant's latest annual environmental report reveals that the company's total electricity consumption has reached an insane 43.6 TWh, with a mind-blowing 97% of that power going straight into feeding ravenous data centers. To wash down all those hot microchips, their water consumption also spiked by over a third, gulping down nearly 11 billion gallons.

We can thank the relentless push for Google Cloud, power-hungry AI Overviews in search, and the endless training cycles of Gemini models for this sudden blackout threat. Apparently, generating a paragraph of AI-written slop requires the kind of infrastructure that used to power heavy industrial factories.

But wait, doesn't Google constantly brag about buying 100% renewable energy? Well, yes, but that is mostly a bureaucratic magic trick. While they contract massive clean wind and solar projects, the physical data centers still have to suck dirty coal and gas energy from the grid when the wind stops blowing and the sun goes down.

While operational emissions look great on paper, their supply chain emissions shot up by 25%, largely due to manufacturing hardware in regions still heavily reliant on fossil fuels. This dragged their total carbon footprint up to a staggering 14.5 million metric tons of CO2 equivalent.

The company claims its custom hardware is getting incredibly efficient, with the latest TPU chips being lightyears ahead of older models, and individual Gemini query footprints shrinking significantly. Unfortunately, the sheer explosion in the number of users asking AI to write emails they could have written themselves completely wipes out any engineering wizardry.

The dream of clean, guilt-free AI is crashing hard against the reality of physics. As tech giants desperately eye nuclear reactors and fusion energy to keep the lights on, the race for artificial intelligence is rapidly becoming a race to see who can deplete the planet's resources the fastest.

Source: Google