Have you ever wondered what the difference is between asking ChatGPT the answer to your homework question and asking Google? Turns out, environmentally, the difference is vast. Researchers estimate that a single ChatGPT query consumes between five and ten times more electricity than a Google search. This article outlines the negative impacts of Artificial Intelligence (AI) on the environment, how the implications of AI on climate change depends on its usages, policies implemented currently, and the next steps for countries.
AI is a term used to describe technology that can process information and almost mimic human thinking. AI is undoubtedly present in our everyday lives. From language models like ChatGPT, to meeting notetakers like Otter.ai, and image generators like DALL-E. To train and run AI, physical data centers need to be built. These data centers require an immense amount of electricity which lead to greenhouse gas emissions, extreme amounts of water usage to cool the hardware, and they produce harmful electronic waste.
At the beginning of its life cycle, AI needs to be trained by feeding data into algorithms to refine their responses– this is an incredibly energy-consuming process. Noman Bashir, postdoc in the Computer Science and Artificial Intelligence Laboratory (CSAIL), argues that training generative AI can require seven to eight times more energy than a typical computing workload. For instance, according to Grist, “researchers estimated that the training of ChatGPT-3, the predecessor to this year’s GPT-4, emitted 552 tons of carbon dioxide equivalent — equal to more than three round-trip flights between San Francisco and New York.” Bashir continues to say, “Total emissions are likely much higher, since that number only accounts for training ChatGPT-3 one time through. In practice, models can be retrained thousands of times while they are being built.” This is significant because ChatGPT is only the beginning to the many more AI models following in its footsteps – each requiring training and emitting greenhouse gases. Furthermore, demands for new AI applications are exponentially increasing. Because of this, companies release new models frequently, which consume more energy for training due to the fact that they attain more capabilities. Consequently, this leads to AI models inheriting a short lifespan; in other words, all of the energy used to train an AI model is wasted once a new AI model is created to be used.
One major concern with the usage of AI is the lack of transparency surrounding its environmental impacts, especially when concerning the emissions from its billions of users every day. Grist states that, “no clear data exists on exactly how many emissions result from the use of large AI models by billions of users.” This is problematic because without this data, it is difficult to pass policies arguing the negative consequences of AI surrounding the general public. Not only that, but without this information, AI users may not be convinced of the environmental impacts of their own AI usage. Adding on, Associate Professor of industrial ecology and sustainable systems at the Yale School of the Environment Yuan Yao, affirmed, “we need transparent, robust methods to assess AI’s environmental impacts. Without accurate quantification, it is impossible to mitigate and address these challenges effectively.” Professor Yao’s comments underscores the importance of gathering quantifiable data to measure AI’s environmental impacts.
Not only does AI consume an incredibly massive amount of energy, the AI data centers also require an extensive amount of water. This is because as hardware is used for training, deploying, and fine-tuning generative AI models, it heats up. Therefore, chilled water is required to absorb heat and cool the hardware. MIT declares that for each kilowatt hour of energy a data center consumes, it is estimated that it requires two liters of water for cooling. For context, Business Energy UK estimates that ChatGPT consumes 39.98 Million kWh per day: using this number leads to 79.96 million liters of water used for cooling data centers for ChatGPT alone. However, Business Energy UK uses numbers calculated from The Washington Post who stated that a 100-word ChatGPT response consumes 519 millilitres of water and 0.14 kWh, leading to a calculation of 148.28 million litres of water per day. The consequences of this include straining local water supplies, disrupting surrounding ecosystems, and damaging biodiversity.
The creation of the hardware devices within data centers alone requires rare earth elements which are often mined in environmentally destructive ways, causing soil erosion and pollution. For instance, these include cobalt, silicon, gold, to begin with. Furthermore, the indirect emissions because of AI include the production, transport, maintenance, and disposal of these hardware components. Not only does the creation of these data centers harm the environment in a materially negative way, but also it impacts the environment while it is running. The data centers produce electronic waste, which often contains hazardous chemicals such as mercury and lead. Furthermore, many of the electronics within the data centers are not properly recycled, seeping these dangerous chemicals into the environment. If not disposed of properly, mercury leads to severe impacts on the environment, especially within marine ecosystems, where it impairs reproductive success and increases mortality rates. Additionally, lead contaminates soil, water, and air, decreases biodiversity, reduces plant and animal growth, and creates neurological effects in vertebrates.
The usage of AI isn’t all negative, though. It depends on who is using it and for what purpose. Climate scientists are taking advantage of AI’s data processing capabilities. For example, Grist shows that by processing large amounts of data, “AI can help create climate models, analyze satellite imagery to target deforestation, and forecast weather more accurately.” This is a major success for climate scientists because without AI these actions would be far less efficient. This is only the beginning of how climate scientists could use AI beneficially. Other examples include improving performance of solar panels, monitoring emissions from energy production, and optimizing cooling and heating systems.
Unfortunately, however, AI is also being taken advantage of by the oil and gas industry to improve the production of fossil fuels. Microsoft, Google, and Amazon hold profitable AI contracts with oil and gas companies such as ExxonMobil, Schlumberger, Shell, and Chevron.
Overall, AI boasts a massive range of problems to our environment: from electricity and water usage to hazardous waste. However, it is up to the user to decide whether the benefits of using AI outweigh the costs, as climate scientists use new technologies, such as AI, to tackle climate change.
Sources:
Bashir, N., Donti, P., Cuff, J., Sroka, S., Ilic, M., Sze, V., Delimitrou, C., & Olivetti, E. (2024). The Climate and Sustainability Implications of Generative AI. An MIT Exploration of Generative AI. https://mit-genai.pubpub.org/pub/8ulgrckc/release/2
Hu, A. (2023, July 6). The overlooked climate consequences of AI. Grist. https://grist.org/technology/the-overlooked-climate-consequences-of-ai/
Ren, S., & Wierman, A. (2024, July 15). The Uneven Distribution of AI’s Environmental Impacts. Harvard Business Review; Harvard Business Publishing. https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts
UN Environment Programme. (2024, September 21). AI Has an Environmental problem. Here’s What the World Can Do about that. UNEP. https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about
Wright, I. (2025, February 17). Business Energy UK. Business Energy UK. https://www.businessenergyuk.com/knowledge-hub/chatgpt-energy-consumption-visualized/
Yao, Y. (2024, October 10). Can We Mitigate AI’s Environmental Impacts? Yale School of the Environment; Yale University. https://environment.yale.edu/news/article/can-we-mitigate-ais-environmental-impactsZewe, A. (2025, January 17). Explained: Generative AI’s environmental impact. MIT News; Massachusetts Institute of Technology. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117