Estimate read time: 5 minutes

Introduction

Using artificial intelligence (AI) has environmental implications. Avoiding AI in the workplace or educational setting is difficult and, perhaps, career impacting, but wanting to be kind to the environment also needs to be considered.  

What?

Training and using artificial intelligence require a large amount of electricity, water, and computer hardware. It also produces vast quantities of greenhouse gases and electronic waste. Some statistics for thought:

Electricity to train AI can release hundreds of tons of carbon dioxide.  There are many tech developers training their own AI systems. 

🏭Estimates suggest that a single AI prompt can generate 4-5 times more carbon dioxide than a single internet search engine query. (Stokel-Walker, & Van Noorden, 2023).

AI requires more powerful computer systems and new data centers are being built world-wide to accommodate the increased demand. These require huge complex electronics supply chains resulting in massive environmental footprints.  

⚡Estimates suggest that a single AI prompt can require 10 times the electricity of a single internet search engine query. (Mearian, 2024).

Water withdrawal and consumption needed to cool AI systems during training and equipment production are massive.  

💧Estimates suggest that producing 10-50 medium sized AI responses can consume 500 mL of fresh drinking water.  Training GPT-3 alone consumed 5.4 million liters of water (mostly potable). (Li, Yang, Islam, & Ren, 2025). 

So What?

We have an ethical and global responsibility to consider the environmental impact of AI use.  The United Nations Sustainable Development Goals (UNSDGs) lists clean water and sanitation (#6), affordable and clean energy (#7), responsible consumption and production (#12), climate action (#13), life below water (#14) and life on land (#15) as key considerations for a global future.

When it comes to using AI, start by asking yourself the following questions; (Simpson, 2025)

Purpose:

Why am I using AI? What's the purpose? Is this necessary or for fun? Are speed/resources critical? Am I trying to create/do something I couldn’t on my own? Has someone already done this? Is this the right tool for the task? Is this aligned with my pedagogy/ethics?

Reusability:

What is the reusability of the material I am creating with AI? Is it one-time use or multi-use? Am I saving/storing generated content? Am I labelling AI generated materials?* If I'm asking my students to use AI, will their generated content be reused?

*Currently, AI generated content can’t be labelled as Creative Commons.

Impact:

Who benefits from me using AI? Who benefits the most? What are the costs for me using AI? Who is bearing the cost? Will it solve a significant problem? Is a company benefiting off the data I put in for training? Am I giving up intellectual property rights? Am I willing to use 2 cups of clean water to do this? Whose 2 cups of clean water am I using? If I had to pay for each prompt, would I use it as much as I am? 

Now what?

How you consider and answer the questions above will impact how you plan to use AI moving forward. This is the responsibility of being a tech steward. Tech stewardship is about making purposeful, responsible, inclusive and regenerative discussions around the creation and use of technology. It's okay to pause on our current path, orientate ourselves, explore new perspectives, deliberate options and move forward on the path in a new trajectory. (Tech Stewardship, 2025)

For those of us who, for whatever reason, need/want to use AI on a regular basis, consider these opportunities to lower the environmental impact. (Simpson, 2025) (Ten Tije, 2024)

• ⚙️Use the right tool and/or right AI model for your needs.  Maybe what you need is available through an internet search.  Maybe there's a tool that doesn't have AI embedded into it.  Maybe there's smaller AI model or an AI model that runs locally on your computer.
• ♻️Reduce, reuse and recycle AI responses. Consider using what already exists, check open educational resources (OER), work in a group to reduce AI usage.  Reuse previously outputs from previously generated AI responses (e.g. save your AI responses for future use).
• 📏Limit output length. Be prompt precise.  Add specifics to your prompt of what you want and don't want in the response.  Put a length limit that fits your purpose.
• 🧑‍🤝‍🧑Group multiple questions/tasks into a single prompt. Ask for a response for each question in a format that works for you rather than multiple individual prompts.

For more details, examples and explanations as to how these suggestions reduce the environmental impact, see 5 Practical API Techniques to Lower Your AI Environmental Footprint - tilburg.ai (Ten Tije, 2024). To read more about Tech Stewardship opportunities at Georgian College, see Staff News. (January 2025 article)

References

Li, P., Yang, J., Islam, M.A., & Ren, S. (2025). Making AI less "thirsty": uncovering and addressing the secret water footprint of AI models. arXiv:2304.03271v4. 

Mearian, L. (2024, September 24). Is the rise of genAI about to create an energy crisis?. Computerworld. 

Simpson, E. (2025, February 7). Climate Conscious AI Use - Wrestling with Environmental Impacts. BC Campus FLO Friday. 

Stokel-Walker, C. & Van Noorden, R. (2023). What ChatGPT and generative AI mean for science. Nature, 614(7947), 214-216. 

Tech Stewardship. (2025). AI Stewardship Practice Program. Tech Stewardship Practice. 

Ten Tije, M. (2024, September 13). 5 practical API techniques to lower your AI environmental footprint. Tilburg.ai. 

Small aspects of this article were adapted from GenAI Quickstart: Foundations for Faculty: Responsible Use Considerations (2024) by Concordia University Library, eConcordia, and McGill University Libraries and is licensed under Creative Commons Attribution 4.0 International.