By: Rose Kores
Edited By: Maddie Miele
The dominance of Artificial Intelligence (AI) technologies in recent years has contributed to advancements in various workforce areas including the healthcare, finance, and transportation sectors. It has enhanced cybersecurity, streamlined data analysis, and improved automation and manufacturing efforts, in addition to making rather mundane tasks such as data sorting more efficient for workers. Despite its benefits, AI has contributed to serious environmental harm through its exorbitant, unsustainable use of energy and water which jointly increase pollution due to carbon and toxic waste emissions. In fact, a 10-message conversation with AI platforms, such as ChatGPT, requires an entire bottle of freshwater to generate responses—a reality that many users must come to terms with to mitigate further environmental harm.1
Demystifying Data Centers
Though the first data center was established at the University of Pennsylvania in 1945, the rise of AI technology has catalyzed the construction of data centers.2 Researchers posit that the energy needs of data centers increased by 2,653 megawatts between 2022 and 2023 due to drastic increases in AI usage. Similarly, electricity usage in data centers around the world collectively surged to 460 terawatt hours in 2022, making data centers the 11th largest energy user.2 Data centers are large concrete buildings sprawled across countless acres of land in a multitude of U.S. states and regions around the world. They contain the computer systems and data servers that power and support AI platforms, relying on substantial infrastructure, specialized hardware, and cooling systems to prevent overheating. According to Smithsonian Magazine, “water that is heated by computers is moved to massive cooling towers on top of a data center, and then is circulated back in. A data center’s direct water consumption is attributed to the water that evaporates during this process. This water loss is then left to the whims of the water cycle.”3 These powerful systems are cooled through methods such as air or liquid cooling, expelling the heat outside of the data centers. For instance, one kilowatt hour of energy requires two liters of chilled water to cool these massive systems.2 Consequently, the heat and water that is released from this process tend to amplify greenhouse gas emissions and pollute local water systems, which are significant environmental ramifications produced by these centers’ current cooling systems.
Unsustainable Energy Reliance
By 2026, AI usage is projected to consume over 6 percent of the United States’ total electricity reservoir. The training process for an AI language model alone requires thousands of megawatts of electricity which emits 552 tons of carbon. Put simply, such emissions are the approximate equivalent of the carbon emissions generated by hundreds of U.S. households in a given year.4 A recent illustration of this is Meta’s data center in Wyoming which is projected to utilize more power than every household in the state combined.5 Aside from surpassing the amount of energy consumed by many households, data centers pose direct effects on households because they often disrupt power grid infrastructure, which in turn shortens the lifespan of many household appliances and blocks access to power and/or internet-reliant structures such as cell phone towers.3 People must consider whether supporting energy-intensive AI platforms is worth the strain on their appliances and the availability of their internet and power systems—a burden that will grow as additional data centers are constructed. If it was common knowledge that a single AI chat query consumes five times more electricity than a standard Google search, perhaps people might begin to reconsider their dependence on and fascination with AI.2
Exorbitant Water Usage
On average, data centers across the United States consume about 449 million gallons of water per day, adding up to roughly 163.7 billion gallons annually. Since less than one-third of data center supervisors and operators track their centers’ water consumption, it is difficult to ascertain precisely how much water these centers consume or how much they will require in the future.1 One study described data centers as being a ‘giant soda straw’ due to their ability to drain entire lakes just to fuel their processing systems. The study explains that data centers in Texas are expected to use 49 billion gallons of water in 2025, a number that is expected to grow to 399 billion gallons by 2030.5 This dramatic increase in water usage–350 billion gallons to be exact– would be the equivalent of utilizing 16 feet of water from the largest reservoir in the U.S., Lake Mead, which spans 157,000 acres.5 Given that freshwater is a finite resource, the extensive use of water to sustain nonessential facilities carries serious consequences for both human communities and the environment. Water that would otherwise be available for drinking and irrigation is displaced by the demands of these data centers. The problem is intensified when data centers improperly dispose of electronic waste, contaminating nearby water systems with hazardous pollutants such as lead and mercury.6 Unless the water demands of data centers are addressed, their growth will continue to strain critical resources and threaten the well-being of the environment and human populations.
Recommendations
Despite their decades-long presence, the recent acceleration in data center development and utilization has reached unprecedented levels. As a result, more research is needed to develop personalized and sustainable solutions that meet the needs of different regions and centers. Nonetheless, immediate action must be taken to begin to address the environmental damage that these centers have caused. Researchers should take the first step in implementing several systems-based changes to reform the use and development of AI language training programs. Instead of generating new language models from scratch, researchers can design domain-specific, customizable models that can be tailored to their different fields or projects.7 Advancements in data centers’ hardware, such as the adoption of neuromorphic chips, can save substantial amounts of money. Similarly, converting data centers’ energy systems to renewable energy sources such as wind and solar has the potential to alleviate fossil fuel emissions and other pollutants.7 Another innovative strategy is to distribute AI computations across multiple time zones, reducing pressure on energy and water resources. This approach also enables workloads to coincide with periods of peak renewable energy availability, which is particularly important given that solar power is only accessible during daylight hours.7
Nations at the forefront of data center expansion, such as the United States, would benefit from following the example of regions like Germany that are actively advancing eco-friendly data center solutions. Germany adopted the Energy Efficiency Act, which imposes strict energy efficiency standards requiring high-consumption companies to implement energy or environmental management systems designed to track performance and energy demand, and to make improvements based on these measurements and analyses.8 The Act also enforces legally binding renewable energy targets for all data centers, requiring that by 2027 their operations be powered entirely by renewable energy sources.8 Furthermore, the Act requires data centers to reduce and reuse waste heat and provide this heat to municipal heat suppliers and networks.8
The Path Forward
While change will not happen overnight, even modest reductions in energy and water consumption, along with improved waste management systems, can meaningfully improve environmental health. As society grows increasingly more technological, data centers are unlikely to diminish in presence or influence. Therefore, as these facilities continue to expand, it is essential that we anticipate their rising demands and proactively address their environmental impacts. By developing domain-specific training models, transitioning to renewable energy sources, distributing AI computations across time zones, and implementing policies, such as the Energy Efficiency Act, we can enjoy the advantages of AI technology while protecting the environment.
Works Cited
-
- Yañez, Miguel. 2025. “Data Centers and Water Consumption.” Environmental and Energy Study Institute. https://www.eesi.org/articles/view/data-centers-and-water-consumption.
- Zewe, Adam. 2025. “Explained: Generative AI’s Environmental Impact.” MIT News. https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117.
- Chen, Amber X. 2025. “A.I. is on the Rise, and so is the Environmental Impact of the Data Centers That Drive it.” Smithsonian Magazine. https://www.smithsonianmag.com/science-nature/with-ai-on-the-rise-what-will-be-the-environmental-impacts-of-data-centers-180987379/.
- Ren, Shaolei, and Adam Wierman. 2024. “The Uneven Distribution of AI’s Environmental Impacts.” Harvard Business Review, (July). https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts?utm_medium=paidsearch&utm_source=google&utm_campaign=domcontent_bussoc&utm_term=Non-Brand&tpcc=domcontent_bussoc&gad_source=1&gad_campaignid=20702632551&gbraid=0AAAAAD9b3uRF6dLBBj.
- Gorey, Jon. 2025. “Data Drain: The Land and Water Impacts of the AI Boom.” Lincoln Institute of Land Policy. https://www.lincolninst.edu/publications/land-lines-magazine/articles/land-water-impacts-data-centers/.
- “AI Has an Environmental Problem. Here’s What the World Can do About That.” 2025. UNEP. https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about.
- Kandemir, Mahmut. 2025. “AI’s Energy Demand: Challenges and Solutions for a Sustainable Future.” Pennsylvania State University Institute of Energy and the Environment. https://iee.psu.edu/news/blog/why-ai-uses-so-much-energy-and-what-we-can-do-about-it.
- Haas, Gabriele, and Christoph Papenheim. 2023. “How Germany’s Energy Efficiency Act will Impact Data Center Operators.” Lexology. https://www.lexology.com/library/detail.aspx?g=56389f61-4fe8-4711-a0dc-6a476927dd9e.
Author Bio
Rose Kores is a first-year Master of Public Administration (MPA) candidate concentrating in Social Policy, with a particular interest in state and electoral politics. She earned her Bachelor of Arts in Political Science from Belmont University, where she gained experience in state and local government, working at the intersection of legislative affairs and constituent services. In the long term, Rose aims to enter the administrative and regulatory law field to merge her passions for law and public policy as a federal legislative strategist.
