In an increasingly interconnected world, it is difficult for technology users to realize the far-reaching consequences of the computing software they use in their daily lives. Open AI’s popular online chatbot, ChatGPT, is trained to respond in detail to user input. The versatility and convenience of this platform has led to frequent usage in business, education and other areas of people’s lives. However, many fail to understand how generative artificial intelligence programs like ChatGPT can perpetuate water loss across the globe.
Recent advances in AI catapulted ChatGPT as an in-demand tool for work and classroom settings. According to the Pew Research Center, around 67% of U.S. teens report that they have heard about ChatGPT. Of those who have heard of this platform, 70% of teens say it’s “acceptable to use when researching something new.” Additionally, another pole from May 2024 found that a fourth of high school upperclassmen say they have used it on their homework assignments. Even young adults are finding use for this, as 43% of adults 30 years old or younger have used ChatGPT– a statistic 10% higher than last summer.
The numbers indicate such software is more than just a fad – the chatbot has drawn attention in a way few other platforms have. In fact, two months after its release in November 2022, ChatGPT amassed 100 million monthly users– crowning it as the fastest-growing web application until Threads took over seven months later.
Turning up the heat
Though the exact cost of this technology is difficult to quantify, it has a significant impact on our natural capital. Each time an answer needs to be generated, the prompt is sent to a server housed in a large data warehouse that utilizes computing power to generate an optimal response. However, these servers generate massive amounts of heat with each request.
The GH Falcon sat down with Corinne Chang (‘25), Vice President of the Green Hope Envirothon club, to discuss the toll this takes on the environment. “The high usage of ChatGPT requires a huge amount of water to keep the engine cool and prevent the system from overheating,” she said. “A liquid cooling system is one of the best ways to keep the server cool since the operation of the server’s CPU produces excessive heat, and they want to prevent this from damaging the inner system.”
As companies race to capitalize on the ever-growing demand for generative AI, the effects on nearby communities are becoming more apparent. OpenAI has borrowed substantial amounts of water from the watersheds in central Iowa in order to cool their supercomputers. Earlier in the year, the West Des Moines Water Works utility company proposed water allotments for the sixth data center Microsoft plans to build in the area. The company is restricted to 5 million gallons a day, however, West Des Moines Water Works also supplies water for municipalities – much of which is set aside for residents.
Even in the most favorable circumstances, it can be seen that data centers are one of the most water-guzzling facilities in a town. Consequently, data centers in hot regions threaten to use immense amounts of water to keep internal temperatures low. In arid areas, this only adds to the rapid depletion of this invaluable asset. Regarded by many as the “purest source of life,” water is still an exceedingly scarce global resource– with just 3.5% of the water that is on Earth being fresh and safe for us to drink and 68% of that amount locked away in ice and glaciers. A report published by the World Meteorological Organization states that the number of people facing water shortages – 3.6 billion – is projected to rise to 5 billion (or roughly two thirds of the world population) by 2050.
Not just water: AI uses energy, too
In some areas, electricity is used to cool systems, rather than water. According to AIRSYS, a company that sells cooling products for data centers, electric cooling systems account for as much as 50% of the data center’s power. They also have a considerable ability to inflate residents’ power bills and strain electrical grids. The increase in demand for new AI technologies threatens to boost the percentage of our overall power that goes to data centers from 1-2% to 3-4%. Although this value is smaller than other sectors, these centers continue to drive up greenhouse gases in the atmosphere.
Chang commented on how this drains our natural capital, affirming that the resources needed to supply any search engine can quickly become unsustainable. “In general, the usage of energy and power has a significant impact on the environment,” she stated. “They oftentimes relied on unrenewable energies and would increase the carbon emission. Also, materials such as lithium would be harmful to our environment if not disposed of properly.”
However, new research following ChatGPT’s rise reveals how it poses the risk of significantly fueling the fire of climate change. Posting a question in their chat channel requires almost 10 times as much electricity as a Google search. As users around the nation cling to this new research resource, it threatens to speed up the rate of resource depletion tremendously.
Before OpenAI’s chatbots are even available to the public, they require large amounts of energy to train them so that they are able to analyze millions of pieces of data in only a few seconds. It is estimated that OpenAI’s latest model, GPT-4, requires 519 milliliters of water to produce a 100-word statement. This amount correlates to a little more than s standard 16-oz plastic water bottle per use. Moreover, when electricity is used to cool the warehouses, ChatGPT’s daily power consumption is roughly equal to that of 180,000 American households (each running on about 29 kilowatts). Used weekly for a year by just a tenth of working Americans requires 121,517 MWh of energy– an amount equal to the electricity used by the entirety of D.C. households for 20 days.
Innovation comes with a cost
Due to its spike in AI research and production, Microsoft (recent partner to OpenAI) saw an unusual 34% increase in its global water consumption from 2021 to 2022. This equates to around 1.7 billion gallons, or around that of 2,500 Olympic-sized swimming pools. However, like many big tech companies, Microsoft has vocalized their intentions to make their data centers more eco-friendly. Though they pledge good intentions, many still lie far from their goals with little progress made.
While meeting corporate sustainability standards for AI production is a large undertaking, some instead consider AI a tool that could help reduce greenhouse gas emissions – if used wisely. A study by the Boston Consulting Group found that artificial intelligence has the capability to accelerate climate action by possibly reducing greenhouse gas emissions by 5-10% by 2030. Though numerous publications label the energy demands of AI as “obscene,” some say that focusing more on what this technology can accomplish rather than how it is made may be worth a shot.
Yet, Chang explained that it is important to hold the possibilities of AI innovation and recognition of its damage in at least equal regard. “I don’t think the benefits of AI could necessarily outweigh the negatives of excessive water and energy use as water is a finite resource that humans need [on] a daily basis. Advancement is important, but the lives of human beings [are] more important.” she said. “So I think a greater concern should be placed on environmental issues in comparison to advancement.”
When considering students responsibility to think about this before choosing their method of research, Chang recognized the complexity of this decision. “In my opinion, it is hard to make students to have moral obligation as search engines is convenient, and of course, we cannot control everyone’s mindset. On the other hand, it will also be hard for us to rely on large corporations to deal with the problems since it is their source of income.”
“Again, I think search engines, even Google, have negative effects on our environment, so I would say it is unrealistic for people to stop their actions.” she said, “Rather, we should work towards government recognition and regulation to limit it to a reasonable amount of consumption each day.”
While the future of generative AI software is difficult to predict, its increasing commonality in the everyday lives of many is apparent. In doing so, innovations like ChatGPT drains water resources – including the ones that its users drink from.
