Summarizing the Environmental Impact of AI

The author is a Year 11 Student (US equivalent, a Sophomore) at Wycombe Abbey School, High Wycombe, United Kingdom.

This story is part of The 89 Percent Project, an initiative of the global journalism collaboration Covering Climate Now.

You’ve undoubtedly seen ChatGPT being used at least once in your life, if you haven’t used it yourself. It seems to have become a permanent feature in each of our lives, for better or worse. According to a study led by the University of Melbourne, in collaboration with KPMG, 66% of people globally use Artificial Intelligence (AI) regularly in 2025, while in 2023 89% of students admitted to using ChatGPT for homework. In terms of the impact on the environment, this will become a very dangerous problem very quickly if we aren’t more conscious of our actions.

To tackle this issue, the 89% project was launched. The initiative is based on a recent study, in which the UN found that 80-89% of people care about climate change and want their governments to take further action. Unfortunately, this figure is not well-known, and so the people that care about climate change often are unaware that they are part of such a vast majority. The 89% project aims to bring attention to this majority and mobilise it. 

AI has become an umbrella term for technology that can perform tasks which usually require human intelligence, and has been around since the 1950s. However, recently, this technology has advanced rapidly, largely due to an increase in funding, research, and computational power. This has been dubbed an “AI Boom,” spanning from the early 2010s to the present. 

AI has great appeal: in simple terms, it has the ability to detect patterns in data and use knowledge of previous events to predict future outcomes. This makes AI a powerful tool for assisting governments, organisations, and individuals. For instance, the United Nations Environment Program (UNEP) uses AI programs to detect when oil and gas installations release methane, a greenhouse gas. When used responsibly and practically, AI can prove to be crucial in the advancement of humanity.

Photo courtesy of Victor Grigas, used via a CCL license.

However, there are some downsides to AI. If we ignore the other negative impacts — such as how AI has reduced the number of entry-level jobs available, or how, according to an MIT study, AI could be eroding critical thinking skills — and only focus on its impact on the environment, AI’s main flaw is its data centre energy and water usage. 

AI data centers store the infrastructure necessary to train and deploy AI programs. These data centers are designed specifically for AI and Machine Learning (ML) programs, meaning they are built for high-performance AI workloads which a traditional data center would not be able to handle. Essentially, this means that AI data centres are much more complex as they are built for high-performance workloads.

To provide energy to these complex systems, ordinarily, fossil fuels must be burnt in order to produce electricity. The burning of fossil fuels in itself also produces greenhouse gases which trap heat, raising the temperature of the earth. In addition, in 2024, the International Energy Agency (IEA) found that AI data centres consumed around 415 terawatt hours, or 1.5% of global electricity; a predicted 23 gigawatts of energy will be burned to power AI in 2025. 

Editor’s note: For additional YEPT stories about AI, see: Digital Footprints: AI’s Environmental Toll (Iowa), Environmental Impacts of AI (Indiana), AI: Climate Problem or Solution? (Ohio), and Your Convenience Or Our Future? (Indiana).

They also use vast amounts of critical minerals, such as lithium, cobalt and rare earth minerals, used in the hardware of servers in data centres. These minerals are mined at the expense of exploited workers, and at the cost of an already struggling environment.

AI data centres use almost inconceivable amounts of water, an already scarce resource. Water is used in cooling systems in data centres to absorb heat and lower the temperature of AI servers: without this they malfunction and shut down, leading to a loss of data. They only use freshwater for this, which makes up 3% of the world’s water, and only 0.5% of the world’s water is both accessible and safe for human consumption. Bloomberg found that about two thirds of new data centres built or in development since 2022 are in places already struggling with high levels of water stress in the US. If that wasn’t shocking enough, on average, half a litre of water is used to cool AI data centre servers for a single ChatGPT user session.

Moreover, they are strikingly expensive. Globally, around $3 trillion will be spent on AI data centres between 2025 and 2029 — approximately the value of the French economy in 2024, according to the BBC. McKinsey predicts that, by 2030, $5.2 trillion must be invested into AI data centres in order to meet the demand for AI globally. To put this into perspective, this is more than ten times Elon Musk’s net worth in October 2025.

Finally, AI data centres produce dangerous electronic waste (e-waste). The term e-waste refers to electronic devices that have been thrown away. Because AI is advancing so rapidly, devices are replaced frequently to keep up with the latest models. Even if not replaced, these devices only have lifespans of 2 to 5 years before they must be disposed of. These substances often contain hazardous substances such as mercury and lead, and can be harmful to human health as well as the environment if devices are not disposed of properly. 

The AI industry is and will continue to keep expanding and developing; there is no doubt that AI will become even more accessible to us in the near future. However, it is critical to question whether we can continue to use AI at such a large scale while trying to preserve what is at risk. The answer to that question must arrive quickly, before it is too late.

Summarizing the Environmental Impact of AI © 2025 by Youth Environmental Press Team is licensed under CC BY-NC-ND 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-nc-nd/4.0/