August 8, 2025

AI’s Dirty Secret

The AI Dream Meets the Carbon Reality

Remember when Google, Amazon, and Microsoft painted themselves green? Net-zero by 2030, 100% renewable energy, tech saving the planet. Lovely stuff. Then ChatGPT fever hit, AI models ballooned, and the cloud suddenly became a lot heavier.

Google’s emissions increased by 11% last year, Amazon’s by 6%, while Microsoft remained 10% above 2021 levels. Meta hasn’t even shown its 2024 numbers yet. The culprit? AI data centres guzzling power like a festival crowd at a free beer tent.

(Source: Claire Brown, The New York Times, via Environmental Health News)

Why AI Eats So Much Juice

Training and running large AI models is not a polite nibble on the grid. It’s a four-course banquet with extra dessert. Every chatbot reply, every AI-driven search result, spins up racks of GPUs, each slurping electricity primarily generated from coal and gas.

Analysts now predict AI data centres could account for up to 12% of U.S. electricity use by 2028. That’s not a typo. And because recent policy changes have scrapped wind and solar tax credits, the extra demand will likely be met by fossil fuels.

Efficiency Tweaks Won’t Save Us

Yes, MIT has proven that you can cut AI’s emissions by 70% if you force it to answer more concisely. Nice party trick, but shaving the foam off your cappuccino won’t offset the fact that you’re now drinking ten cups a day.

The problem isn’t just inefficiency, it’s scale. Big Tech is building server halls the size of small cities. Utilities respond by firing up old gas plants, because clean energy isn’t scaling fast enough to keep pace.

The Water We Don’t Talk About

Electricity isn’t AI’s only vice. Data centres need cooling, and lots of it. In drought-prone regions, this means pulling millions of litres from stressed rivers and aquifers. If you’re in Greentech or sustainability, that’s a PR nightmare waiting to happen.

What This Means for You

If you’re leading a scale-up or building AI-heavy products, you’re not immune. The infrastructure choices you make now could lock you into a future with dirty energy and make you a target for ESG scrutiny.

Key takeaways:

• Factor sustainability into your tech roadmap, not as an afterthought but as a cost of doing business.
• Choose cloud partners with verifiable renewable energy commitments (not just nice marketing decks).
• Track your own AI energy footprint and water use. You can’t manage what you don’t measure.
• Consider efficiency at the design stage, leaner models, smarter inference, and fewer vanity features.

Conclusion: The AI gold rush is real, but so is the carbon tab. Big Tech’s net-zero pledges are slipping, and if the giants can’t hold the line, smaller players have to be twice as smart. In the AI era, sustainability isn’t just a virtue signal; it’s a competitive edge.

Back to news