If there’s a task that needs doing you can bet that there is an AI tool that can help you do it. From writing your own articles, editing photos, generating imagery, organising your schedule, contacting your leads. AI is everywhere. Used correctly AI can be great at helping you reduce your impact on the environment, but without care they can significantly increase your carbon footprint.
We’re seeing more online trends with AI and less productivity. Nobody needed to see the doll version of the professional self and yet thousands of us jumped on board to see what accessories our working-self would come with. The outcome? A significant increase in universal energy and water consumption.
In this article we’re going to talk about how AI is affecting our climate but more importantly demonstrate just how much energy is being consumed from basic AI actions. The impact of your ChatGPT search, the energy storage for all those servers, other emissions and resources that are depleted from AI image creation and more.
Want to know if AI is harming your carbon footprint? Keep reading.
Table of Contents
So, how much energy are we talking about?
Let’s break it down. Every time you prompt an AI like ChatGPT, it is not just your device doing the work. Your query triggers vast data centres full of high-powered servers that are constantly running and generating heat, which then needs to be cooled using more energy.
According to research from the University of Massachusetts Amherst, training a single large AI model (such as GPT-3) produces around 284,000 kilograms of CO₂ emissions. That is about five times the lifetime emissions of an average car, including its fuel.
And it does not stop after the training phase. Everyday use, or inference, still draws energy. A single ChatGPT query is estimated to consume between 0.3 and 1 watt-hour of electricity. That is up to 10 times more than a Google search, which uses around 0.0003 kilowatt-hours per search. While this might seem minor, it adds up quickly when you consider that AI systems handle hundreds of millions of queries every day.
AI image generation? Even more resource-heavy
Text queries are intensive enough, but image generation takes things to a whole new level. Using image generations tools can consume more than 3 kilowatt-hours for every 1,000 images generated, according to research by the University of Cambridge.
To put that in perspective, that is enough electricity to fully charge 270 smartphones or run a laptop for 100 hours. Multiply that across the millions of AI-generated images being created for fun, branding, or novelty, and the energy usage becomes difficult to ignore.
Water usage is another hidden cost. A study from UC Riverside and the University of Texas at Arlington found that training GPT-3 used approximately 700,000 litres (or 184,000 gallons) of clean water, mainly for cooling servers. Even something as small as 20 to 50 prompts to ChatGPT can consume around half a litre of fresh water, depending on when and where the servers are operating.
The environmental cost of storage
Let’s not forget that all your AI-generated documents, images and videos need to be stored. Cloud storage uses real-world infrastructure, and that infrastructure uses electricity. The International Energy Agency (IEA) estimates that data centres already account for 1 to 1.5 per cent of global electricity consumption, and this could rise to 8 per cent by 2030 if current trends continue.
How to reduce your AI carbon footprint
The good news is that small, conscious changes in how you use AI can significantly reduce your impact.
1. Use AI more intentionally
Before opening your favourite AI tool, ask yourself if it is really necessary. Do you need help drafting a blog post, or are you just playing around? If just 10 per cent of users skipped one unnecessary interaction each day, the energy savings could reach hundreds of megawatt-hours each year.
2. Be concise
Shorter prompts mean fewer computations, which saves energy. You also tend to get better results more quickly, making your usage more efficient overall.
3. Limit image generation
Try not to generate dozens of AI images unless you genuinely need them. One or two visuals for work or education is reasonable. Fifty fantasy-style avatars of yourself is probably not.
4. Choose greener providers
Some AI companies are making strides toward sustainability. Google and Microsoft, for example, have pledged to run their data centres on 100 per cent renewable energy by 2030. When choosing AI tools, look for companies that are transparent about their environmental impact and are actively reducing it.
5. Clear out digital clutter
Unused files still consume energy through storage. Delete unnecessary AI-generated content from your cloud storage. A tidy digital workspace is good for you and better for the planet.
Is AI an eco-nightmare?
AI is not inherently bad for the environment. In fact, it has the potential to support climate science, predict weather patterns, optimise energy use, and more. But casual and excessive use has consequences, especially when scaled across millions of users worldwide.
If you are serious about reducing your carbon footprint, it is worth taking a closer look at your digital habits. Being more mindful of how and why you use AI is a simple but effective step towards a lower-impact lifestyle.
So next time you are tempted to generate 30 AI versions of your pet or ask ChatGPT what kind of pasta your personality would be, maybe take a second to consider the real-world cost behind the screen.
