Burning Down The House: The Carbon Footprint Of Your AI Addiction!

Burning Down The House: The Carbon Footprint Of Your AI Addiction!

The digital age has brought us many conveniences, chief among them the ability to generate endless content with a few keystrokes. But behind every clever AI-generated headline or image for social media, piece of copy or report, lies a rather inconvenient truth: it’s all consuming a startling amount of the planet’s resources. While we pat ourselves on the back for going paperless, we’ve quietly embraced a new, deeply thirsty, and power-hungry habit.

At the heart of this issue is a simple matter of scale. To create content, AI models must first be trained and then run on vast farms of servers. These are not your average office computers; they are powerful, purpose-built machines demanding colossal amounts of energy. Training a single large model, such as GPT-4, can require an estimated 10 to 100 megawatt-hours of electricity [^1, ^2]. That’s enough to power a small town, just so an algorithm can perfect its ability to write your punchy prose. 

It seems we’ve successfully outsourced our creativity to a process that is burning down our collective home. The International Energy Agency predicts that data centres, fueled by our insatiable appetite for AI, could soon account for up to 4% of global electricity consumption [^3, ^4]. Goldman Sachs, meanwhile, projects a tidy 165% increase in data centre power demand by 2030 [^5]. A truly impressive, if slightly alarming, growth story.

And then there’s the water. In a world of increasing water scarcity, our data centres are, for lack of a better term, incredibly thirsty. They require immense quantities of water for their cooling systems, which prevent the servers from overheating and melting into a very expensive puddle. It’s been estimated that a single, brief conversation with a chatbot can consume approximately half a litre of fresh water, while training a large model can consume hundreds of thousands of litres [^8, ^9]. In 2024, Google’s data centres alone consumed nearly 6 billion gallons of water, with the increase driven specifically by AI advancements [^7]. The irony isn’t lost that some of these facilities are located in water-stressed regions, allowing us to generate a witty tweet at the expense of a very real, and very dry, resource [^8, ^9].

While some might argue that AI’s efficiency could ultimately help us address climate change [^10], the more immediate reality is that the tools we use for our digital convenience are contributing to the very problem they might one day help to solve. For now, every plan, image, and social media post generated by AI adds to a tangible environmental footprint. It seems that machine learning is doing more than just learning; it’s also quietly running up a considerable utility bill, both for us and for the planet. Perhaps it’s time to revisit the skills and relatively low consumption of the professional, creative, human being.

NB: I’ve read many polemic pieces and wondered where they get their background facts from; those funny [^] dotted around the piece refer to the articles I have utilised and they are linked below, should you wish to read further.

[^1]: Built In. “AI Energy Consumption: Is It a Problem?”
[^2]: Innovation News Network. “New TUM training model slashes AI energy consumption.”
[^3]: Goldman Sachs. “AI to drive 165% increase in data center power demand by 2030.”
[^4]: Computing UK. “IEA: AI’s energy demand to quadruple by 2030.”
[^5]: Chancery Lane Project. “Artificial intelligence and energy usage: crisis or catalyst?”
[^6]: Water Unite. “Rising Water Demand from Data Centres: Addressing AI’s Water Impact Through Innovation.”
[^7]: Anadolu Ajansı. “Google data centers used nearly 6B gallons of water in 2024.”
[^8]: Investopedia. “The Hidden Cost of AI: How Data Centers Are Draining Water Resources and What It Means for Investors.”
[^9]: UCI Library Research Guides. “Generative AI and Information Literacy: Other Ethical Challenges with AI.”
[^10]: LSE. “New study finds AI could reduce global emissions annually by 3.2 to 5.4 billion tonnes of carbon-dioxide-equivalent by 2035.”

  • Paul Richardson

    Paul Richardson FCIM MCIPR is a marketing and PR specialist at Vividink, mainly working in the events and exhibitions industry.

We need your help

We believe the events industry deserves a central, independent space for news and resources that isn't hidden behind a paywall and is free of banner ads. At just £10 a month, you can help to ensure EPN Journal remains ad free and accessible to everyone in the event industry. We use Stripe for security and you can cancel any time.