The Environmental Impact of Data Centers and Artificial Intelligence

Casual social media scrolling, internet browsing, and streaming music are some of university students’ favorite pastimes, but not many people realize that their relaxation has consequences for the planet. Currently, there are 8,612 data centers with servers computing, storing, and distributing data around the globe. Data centers are comprised of high-performance computers, hard drives, and large infrastructure, with the largest data center spanning almost a million square meters. The United States accounts for 36.9% of all data centers around the world (Data Center Map, 2025). While this saturation charts the United States as a hub for technological innovation, it poses concerns about the environmental implications. Americans pride themselves on advancements that bring the future to the present, but we must ensure there is a future waiting for us.

Environmental Costs of Data Centers

Artificial intelligence, internet browsing, and other digital tasks are powered by external processing systems in data centers, which require energy. Power plants fuel the data centers while releasing copious amounts of carbon dioxide, a greenhouse gas, into the atmosphere. In 2024, 140.7 megatons of CO2 were produced by data centers, which would require 6.4 gigatons of trees to absorb all the CO2. It would take over 10 million coast redwood trees, the tallest tree in the world, to absorb all the CO2 produced by data centers in one year. In one year, data centers demand 256.6 terawatt-hours of energy, enough to power 24 million households (Guidi, et al., 2024). The statistics behind the energy-intensive mechanism of data centers are concerning and will only accelerate at the rate the world is advancing.

Water is a scarce and sacred commodity that runs through the veins of every person on the planet. It also runs through the veins of a data center, pumping water around the computers to keep them cool. There are three main sources of water: potable water, reused water, and treated effluent, sewage that has been cleaned. Ideally, potable water would be reserved for human consumption and usage, but reused water and effluent often contain contaminants that cause corrosion and bacterial growth in the pipes. While most data centers have a cyclic system to reuse water, it can only undergo a finite number of cycles until it must be discharged. Furthermore, data centers not only require a constant demand for new water but also commonly store 10 million gallons of emergency water in case the main system malfunctions (Ahmad, 2024). These centers are water-intensive, a dilemma that is exacerbated in hot climates. To create the most sustainable data center water system, engineers should construct sites in cool environments, create a cyclic cycle that utilizes treated effluent, and treat the water before it is discharged.

Energy Consumption of AI

One ChatGPT inquiry uses the amount of energy it takes to light a bulb for twenty minutes (Kerr, 2024). However, each inquiry is not equal. There is variation between different tasks and the number of emissions produced by each inquiry.

To define some of these tasks, here are two examples of AI chatbot inquiries and the tasks they require to produce a response:

You ask an AI chatbot, “How many points is a touchdown?”

Text classification: Categorizes the inquiry as “sports.”
Masked language modeling: Recognizes that “touchdown” is supposed to mean “touchdown.”
Text generation: Creates text to tell you the answer.
You upload a screenshot from a football game and ask the AI chatbot, “What is happening on the field?”

Object detection: Identifies the football, players, and yard lines.
Image classification: Labels the objects in the image as “football,” “players,” and “yard lines.”
Text generation: Creates text to describe the image.
Image generation requires the most amount of energy, using 2.907 kWh for 1,000 images, significantly more than the next highest energy-taxing task, image captioning which uses 0.063 kWh for 1,000 captions (Luccioni & Jernite, 2024).

The race for the most sophisticated AI system is the catalyst for the constant construction of new data centers. Despite branding themselves as eco-friendly companies, large corporations are the worst offenders of greenwashing. Google claims it is dedicated toward achieving net-zero emissions by 2030 but their greenhouse gas emissions rose 48% from 2019 to 2024. This rise in emissions is attributed to their Gemini chatbot and overview feature, which require infrastructure and AI training that require energy (Kerr, 2024). Google has repeatedly demonstrated that AI development is their top priority. To ease the public, Google often implements sustainable facades when it introduces energy-intensive systems. Figure 3 displays the image shown on their website for the data center in Council Bluffs, Iowa. It depicts a family of deer roaming outside the data center, trying to convince the audience that their AI advancements can coexist harmoniously with nature. However, the undertone of the image depicts a family of deer in headlights, eerily looking at the people behind a computer screen, knowing their lives are being threatened by environmentally negligent companies. With five years remaining on their net-zero promise, the clock is ticking for Google’s AI aspirations to align with their sustainable claims.

The Equity Issue

As new data centers are being constructed around the globe, an issue of equity arises. Using acres of land to create a data center places a strain on the local community, imposing noise, air, and water pollution on communities with scarce resources. Ironically, data centers can also act as computing systems that replace the jobs of people in the community. In 2022, Meta aspired to construct a massive data center in a small Dutch town called Zeewolde, which is the only city in the Netherlands that produces more renewable energy than it uses fossil fuel energy. Meta was attracted to the renewable energy available for its data center, but it would demand the amount of energy it requires to power all the homes in Amsterdam. Additionally, residents were concerned about the loss of agricultural land, water contamination, and stealing renewable energy from the community. Local activists protested the data center and urged their government to protect its land and people (Hamilton, 2022). Later that year, the project was canceled by Meta. Zeewolde is an excellent model for environmental activists to promote justice, but we must ensure that all communities can voice their opposition against powerful tech companies. Without regulation, underserved communities will be colonized by greedy corporations seeking to replace cultural centers, family homes, and local ecosystems with acres of computers.

Environmental pollution widens the gap between socioeconomic groups because disadvantaged communities do not have the resources to avoid pollution and treat the consequential health conditions (Liao, et al., 2023). To ensure the construction of data centers does not exacerbate environmental and socioeconomic inequities, data centers should be distributed equally around the globe, with special attention to not placing them in locations that experience frequent drought or extreme heat conditions.

As university students, we can join the cause in three ways:

Know the environmental impacts of having a presence on the internet.
Advocate for data centers to contribute to their local communities instead of depleting resources and exacerbating socioeconomic disparities.
Avoid asking AI to generate images.
Bonus) Avoid misspelling text to an AI chatbot or inputting essays or images for it to analyze.

 

Works Cited

Ahmad, Rasheed. “Engineers Often Need a Lot of Water to Keep Data Centers Cool.” Asce.org, 2025, www.asce.org/publications-and-news/civil-engineering-source/civil-engineering-magazine/issues/magazine-issue/article/2024/03/engineers-often-need-a-lot-of-water-to-keep-data-centers-cool#disqus_thread. Accessed 25 Feb. 2025.

Guidi, Gianluca, et al. “Environmental Burden of United States Data Centers in the Artificial
  Intelligence Era.” ArXiv (Cornell University), 14 Nov. 2024, https://doi.org/10.48550/arxiv.2411.09786.

Kerr, Dara. “AI Brings Soaring Emissions for Google and Microsoft, a Major Contributor to Climate Change.” NPR, 12 July 2024, www.npr.org/2024/07/12/g-s1-9545/ai-brings-soaring-emissions-for-google-and-microsoft-a-major-contributor-to-climate-change.

Luccioni, Sasha, et al. “Power Hungry Processing: Watts Driving the Cost of AI Deployment?” Association for Computing Machinery, 3 June 2024, p. Pages 85 – 99, https://doi.org/10.1145/3630106.3658542. Accessed 6 July 2024.

Tracy Brown Hamilton. “In a Small Dutch Town, a Fight with Meta over a Massive Data Center.” Washington Post, The Washington Post, June 2022, www.washingtonpost.com/climate-environment/2022/05/28/meta-data-center-zeewolde-netherlands/.

“Worldwide Data Centers – Colocation and Cloud.” Www.datacentermap.com, www.datacentermap.com/datacenters/. Accessed 24 Feb. 2025.