AI’s Environmental Impact
Written by: Sophia Hernandez
Edited by: Nicole Bowen, Beatrice Bos
Introduction
Put simply, artificial intelligence (AI) is the largest technological advancement made in the 21st century. Not only does it compute and process information at inhuman speeds, but its effectiveness and popularity among all sectors of industry has made it an indispensable tool.
Even today, in AI’s infancy, it can be trained to analyze data with an unmatched efficiency, computing and generating information to reach conclusions. In daily life, we receive infinite assistance with recommendations, analysis, and communication. In nuanced lines of work such as healthcare and scientific research, AI assists with making breakthroughs for disease prevention, treatments, and countless other factors.
It’s projected to contribute 15.7 trillion dollars to the global economy by 2030. One would suppose we would know more about the environmental implications of AI, considering its mainstream usage among companies and individuals. Yet, most don’t, and if they do, they do not share the same sentiments that corporations have.
Reader, let me ask you: do you know about how AI affects our natural world? And if you did, would you shift your perspective on using it in your life?
(NB: alternative phrase: I urge the reader to question how much you know about AI affecting our natural world, and whether this has changed your use
AI Development
AI does not start as the information giant we’ve come to know, but rather a prototype that needs to be trained within a data centre. These data centers act as facilities that house the complex IT infrastructure AI needs to be trained in order to deliver real time results. Essentially, data centres are large buildings that contain technology built to transmit and store large amounts of energy to train and maintain AI models.
These data centres use significant amounts of energy on each and every AI model throughout its training, specifically in the form of carbon dioxide, which is a gas that traps heat in Earth’s atmosphere. The University of Massachusetts, Amherst found that a single AI model releases 626 pounds of carbon dioxide when being trained, which is 56 times moreenergy than a human releases during one year of life.
Why does this matter? The excess carbon dioxide released by AI model training adversely affects the Carbon Cycle, an important process in nature that helps regulate Earth’s temperature and climate. A surplus of this gas results in more difficulty for areas that absorb greenhouse gases, also called carbon sinks, to contain the excess CO2. Because of this, more carbon dioxide will go into our atmosphere, warming the global temperature and oceans in the process.
The Real Cost of Maintenance
AI data centres need lots of energy to maintain themselves, so they tend to overheat. For example, one AI model called Chat GPT, known for their virtual assistance in everyday life, consumes enough energy to power an entire house. This energy surplus requires a cooling agent to regulate heat generated by energy production and transmission, which typically comes in the form of water. Several methods have been implemented, but all follow a similar cycle.
A freshwater source, such as a river or dam, has water extracted from it
The water circulates around to cool the data centre
This water is disposed of or deposited back into the original water source once no longer cool; alternatively, it evaporates into the atmosphere
U.S. Department of Energy
Data centers participate in both water withdrawal and consumption, two often confused but distinct processes. Water withdrawal involves extracting water from a natural resource. Alternatively, water consumption involves utilizing this water, and either causing it to evaporate or return to its original source in an altered state (for example, a different temperature).
Water is a finite resource, meaning that there is a limited amount of it on Earth. Just 0.5% of Earth’s water is suitable for agricultural and municipal use, since salt water cannot be used for these purposes. While it is arguably more abundant than other natural resources such as minerals, its usage in cooling AI data centres exacerbates water scarcity on a global scale. The Organization for Economic Cooperation on Artificial Intelligence found that “…global AI demand {for water} may even require 4.2 – 6.6 billion cubic meters of water withdrawal in 2027, which is more than the total annual water withdrawal of 4 – 6 Denmark,” Essentially, this means that AI is consuming more water than countries can provide, taking away a vital resource from several towns and cities.
Case Study: The Dalles, Oregon
Home to a population of 15,000, residents of this small town near the Columbia River are facing a troubling situation likely to worsen with the increased use of AI.
The Dalles, Oregon boasts a Google data centre that acts similarly to data centres previously described; since it acts as a hub for information, it requires exhaustive amounts of water to prevent overheating. According to the local media outlet Oregon Live, “the facility has nearly tripled its water use in the last five years, and the technology multinational plans to open two more data centers along the Columbia River”
Director of the nonprofit advocacy group WaterWatch John DeVoe predicts adverse consequences of the continued growth of AI usage and related data centres. He believes if data centers’ water use in The Dalles doubles or triples over the next 10 years, it will have detrimental effects on fish, wildlife, and municipal water.
The Dalles, Oregon is only one example of how AI’s water usage can compromise wildlife and public water. With the increase of AI, it will not be the last.
What Can We Do?
There is not a linear solution to this growing issue, but there are individual actions we can take to reduce our AI usage. It starts with knowing forms of artificial intelligence in our daily lives.
Artificial Narrow Intelligence (ANI) - “systems that are designed to perform specific tasks or solve particular problems within a defined scope—they cannot think or make decisions beyond them.”
Virtual assistants (Siri and Alexa)
Recommendation algorithms (media feeds)
Facial recognition
Mapping systems
Reactive Machine Artificial Intelligence (RMAI) - “systems respond to specific inputs with predetermined outputs without the ability to store data or learn from past experiences.
Streaming service recommendations
Traffic and stoplight management systems
Limited Memory Artificial Intelligence (LMAI): “systems that can store and use past data to improve their predictions and performance over time.”
Self driving cars
Customer service bots (online job applications, company chat bots)
Smart home devices (smart fridges, smart showers)
Industrial robotics
Academic assistants (Grammarly, PhotoMath)
Theory of Mind Artificial Intelligence: “represents a future stage of artificial intelligence that aims to understand and respond to human thoughts and emotions.”
ChatGPT
Chatbots (CharacterAI, Chai)
Some of these are unavoidable. We cannot dictate how our traffic lights function, or what appears on our media feeds. However, we can decide whether to install such as academic assistants into our study routines, smart appliances into our homes, and chatbots into our downtime sessions. Power is found in knowing what we can control.
AI is an extremely useful tool. It has the opportunity to completely revolutionize the way we work as a society, from playing a key role in scientific research and discovery to changing how we go about daily tasks. It’s fundamental to be aware of its potential, but also of its limits. The 20th and 21rst century has been characterized by progress in the technological field; however, they’ve also been struck by the climate crisis, a phenomenon that's only going to worsen in upcoming years if things don't change. In light of the current environmental emergency we are facing, we cannot ignore the environmental impact of AI, despite all the great benefits these developments can bring. It’s important to use artificial intelligence with mindfulness and consciousness of its impact: appreciating its potential for positive changes also means recognizing areas in which AI can cause harm, and acting accordingly to promote a responsible use of these technologies.
Further discussed in:
Fawley, S. (2024). 9 Benefits of Artificial Intelligence (AI) in 2024. [online] University of Cincinnati. Available at: https://online.uc.edu/blog/artificial-intelligence-ai-benefits/.
Ren, S. (2023). How much water does AI consume? The public deserves to know - OECD.AI. [online] oecd.ai. Available at: https://oecd.ai/en/wonk/how-much-water-does-ai-consume.
Strubell, E., Ganesh, A. and McCallum, A. (2019). Energy and Policy Considerations for Deep Learning in NLP. arXiv:1906.02243 [cs], [online] 1. Available at: https://arxiv.org/abs/1906.02243.
United Nations. (2025) Artificial Intelligence: How much energy does AI use?. [online] UN Available at: https://unric.org/en/artificial-intelligence-how-much-energy-does-ai-use
Pascual, M (2023). Artificial intelligence guzzles billions of liters of water [online] Available at: https://english.elpais.com/technology/2023-11-15/
Rogoway, M (2023). Google’s water use is soaring in The Dalles, records show, with two more data centers to come [online] Available at: https://www.oregonlive.com/silicon-forest/2022/12/googles-water-use-is-soaring-in-the-dalles-records-show-with-two-more-data-centers-to-come.html
Syracuse University (2025). Types of AI: Explore Key Categories and Uses [online] Available at: https://ischool.syracuse.edu/types-of-ai/