The Environmental Impact of AI: What We Know, What It Means & What You Can Do
AI has become part of everyday work for many people, from quick idea generation to content support and data analysis.
With that growth has come a valid question. What is the environmental impact of using AI and how should we think about it?
How AI Uses Energy
AI uses energy in two main stages. The training stage and the everyday usage stage.
Training
Large AI models are trained using powerful hardware over long periods of time. This is energy intensive but don’t happen as often. If you want to dive into more detail, The UN Environment Programme has a good overview in their report here: UNEP: “AI Has an Environmental Problem”
Everyday Usage
Most of the environmental footprint comes from what happens when millions of people use AI each day. Every prompt you send requires a model to run calculations. The more people use AI, the more energy is required across datacentres. UNRIC has more info on this as well.
In simple terms, training is a one-off cost. Using the model is the ongoing cost.
What Is A Data Centre?
For most of us, we use online technology without really thinking about the mechanics behind how that works. This is important to understanding how AI usage is impacting the environment.
Most of what we do online requires a data centre. They are essentially large buildings with big machines that power the internet. In doing so, they produce a lot of heat and use water to cool that heat. This is where the environmental impact comes in.
AI is essentially contributing to the effect that data centres have on the environment.
What That Impact Looks Like in Real Terms
If most of the impact comes from everyday usage, then we need to understand how AI prompts compare to the other ways we use technology.
A single AI prompt does not have a huge footprint by itself, but it is more energy intensive than most typical digital actions. That means more energy output via the data centre compared to other types of online usage.
AI usage sits somewhere above sending an email and below streaming a video. If you’re using it for image and video generation then that does require far more energy than text.
So while AI is not the biggest digital carbon offender, it is not negligible either and should be treated with the same awareness we apply to streaming, cloud storage or endless content refreshes.
However, it’s hard to measure the exact impact of AI usage.
Why It Is Hard to Measure the Footprint Accurately
Unlike transport or heating, where numbers are clearer, AI’s footprint is harder to pin down for a few reasons.
Different companies train models in different ways
Data Centres run on different energy mixes in different countries
Hardware varies in efficiency
Usage patterns change constantly
Providers do not always publish full lifecycle data
This means most figures you see online are estimates rather than exact measurements.
The Other Side of the Equation
AI is not only an energy consumer. It can also support environmental outcomes when used well. For example:
Monitoring wildlife and ecosystems
Analysing climate data
Reducing wasted resources in supply chains
Assisting research that would otherwise take significant manual time
Helping small teams work more efficiently
It is useful to see AI as a tool with both costs and potential benefits, depending on how it is applied. The Mammal Society produced a great article on how AI can help conservation for example.
What You Can Do
The message I would give is that using AI is the same as anything you do on the internet. I encourage you to be mindful that it has a heavier impact than other types of internet usage, but it does not sit separately to what is largely an issue around data centres - not AI.
Having said that, there are steps you can take to be more intentional in how you use it.
Use AI when it adds real value: not every task needs AI. Save it for work that benefits from clarity or speed.
Choose lighter models where possible: smaller or more efficient models use less energy per request. Or you could create a CustomGPT that uses less energy.
Avoid unnecessary image or video generation: these use far more energy than text. Avoid using it for ‘fun’.
Reuse prompts and workflows: if you need similar outputs, refine your prompts and re-use them rather than starting from scratch each time.
Look for providers with strong sustainability commitments: the energy mix of the datacentre matters more than many people realise.
RaptorBeak’s Approach to Responsible AI
At RaptorBeak, AI is used carefully and transparently. It supports early-stage structure, clarity and ideation, but it never replaces human judgment or the authentic voice of the people I work with.
I avoid high-energy workflows unless they genuinely add value, and I keep the focus on clarity rather than volume. As a member of 1 Percent for the Planet, a portion of revenue goes directly to environmental causes, which helps balance out the digital tools I use.
The aim is simple. Use technology in a way that feels aligned with the values of the people I support.
Final Thoughts
AI has an environmental footprint, just like every digital tool we rely on. The goal is not to avoid it entirely, but to understand its impact and use it with intention. With thoughtful habits and better infrastructure, the footprint of AI can be managed while still offering real benefits to the people who depend on it.
If you have questions about how AI fits into ethical digital marketing, I’m always happy to talk.