AI puts technological tap on Great Lakes, natural resources
Increasing the drain of Great Lakes water to power AI data centers triggers both optimism and concern, yet the full range of its impact remains uncertain due to a lack of transparency from tech companies and reports. As AI usage has increased since 2021, tech companies are aiming to build more data centers near the Great Lakes. Since data centers use high amounts of energy, they need to be cooled down with freshwater. Not only are the Great Lakes an abundant resource, but their climate is relatively cooler than other locations where water is abundant. However, tech companies have been receiving backlash for the lack of transparency regarding just how much water is used to enable AI.
Dr. Gregory T. Kleinheinz, a professor and Director of the Environmental Research and Innovation Center at University of Wisconsin–Oshkosh, feels companies haven’t been forthright with information.
“AI data centers and their owners are not sharing information on water usage and they are making local officials sign NDAs (non-disclosure agreements) so they do not disclose the amount that is being used,” he said. “There needs to be better transparency.”
photo from Shutterstock
In August 2025, Alliance of the Great Lakes published a report regarding the increased use of the Great Lakes’ water for AI. They stated that the use of NDAs, a lack of requirements for reporting water use, and inconsistencies in water management laws all brought forth obstacles for states and local governments in both financial growth and resource management. In 2023, another study from the University of California, Riverside relayed that OpenAI’s GPT-3 claims that running 20-50 queries with 20-50 questions in an AI conversation uses half a liter (500 milliliters) of water. This did not include the vast amount of other user behaviors like single prompts or multi-stage conversations.
A lack of research makes data difficult for this issue, according to West science teacher Logan Martilla
“From the data center side, they are unsure of how much water is truly getting used and only have rough estimates,” he said.
In 2025, Google released the data on how much water their AI systems used. The Google Cloud blog post claimed the “median Gemini Apps text prompt” uses 0.26 milliliters of water, which is approximately five drops. Similarly to the previous study, Riverside did not specify how long a prompt should be if it only uses around five drops of water, nor does it specify about other requests. According to Google’s data, if scaled to match UCR’s study, Gemini uses around 13 milliliters of water per 20-50 queries while Open AI’s GPT-3 uses 500 milliliters. However, there is no standardized measurement for AI water consumption, and the studies mentioned provide limited data to demonstrate how much water is used in practice.
Kleinheinz highlighted the importance of obtaining exact measurements of water.
“Transparency of the amount of water used with measurements and data makes sure the water is returned to our correct waterways in a manner that does not harm the native ecosystem,” he said. “Any Great Lakes water needs to be returned to the Great Lakes.”
Science teacher John Reiland agreed with the returning of water to its source.
“I do believe strongly that water from one watershed needs to be put back into the same watershed,” he said. “The Midwest has 20 percent of the world’s freshwater, so we need to take care of it.”
West freshman Ariel Verdic worries about the self-sufficiency of using freshwater to power data centers.
“If we keep using water to cool down AI, we might fry the systems or run out of water we need to use for cooling it down,” she said.
Although some stress that the data centers are detrimental for Wisconsin water sources, others support the creation and preservation of AI systems.
Freshman Dylan Yang has a positive outlook about the spread of artificial intelligence.
“I think that the water that we are using will come back through the water cycle, so it won’t actually disappear,” he said. “And we need AI because it saves so much of our time. It makes it so that pros can focus on bigger problems and let AI fix smaller problems.”
AI is a prominent and rising technology used for a variety of purposes. Use of ChatGPT, Google Gemini, and chatbots has skyrocketed in recent years, which calls for more water to power it. Though the reality is uncertain, Martilla reasons that the water cycle might not be able to catch up to the speed at which AI must be powered to meet demand.
“They also don’t really have a lot of procedures in place to conserve water or recycle it from the steam produced, so we could be using water much faster than it can condense back and fall as rain,” he said.
Martilla stressed that users need to start being more efficient and sparse with their use of AI.
“People truly do not know how to use AI responsibly energy-wise,” he said. “Large portions of the population are incredibly friendly with AI and add lots of filler words, long questions, and ‘thank yous’ to prompts which cause the servers to compute much more information than what is necessary. This then causes more water to be used to respond to prompts.”
Another potential concern highlighted by Martilla is water pollution, which may rise in potency with diminished water levels.
“There will also have to be constant monitoring on pollutants in the water because the concentration of pollution to water will rise, making the water more toxic as the water level goes down,” he said.
AI has grown as a great influence on the Wisconsin waterways, but the exact impact on natural resources remains unknown. Martilla presents the questions that Wisconsinites need to ask as data center presence grows in the state.
“While tech companies are not being transparent with their information, AI databases still have a significant impact on Great Lakes water,” he said. “Ultimately, it's ‘how much water is this going to use and how is it going to affect the ecosystem of the lakes?’”
by Yui Watanabe
Published October 6, 2025
Oshkosh West Index Volume 122 Issue I