Artificial intelligence, real emissions

Design by Alexandra Kolaya

There is no question that artificial intelligence has infiltrated our daily lives. From Google searches to Netflix recommendations, AI has become nearly unavoidable. Many criticisms of AI focus on creativity, ethics, and academic honesty. But what people may not realize is that with every AI-generated message, image, or search, there’s also an ecological cost. 

At Carrollton, students say they use AI almost daily. 

“I use ChatGPT to help with math problems or to explain a topic step by step, kind of like a tutor,” said one student. 

Another added, “I like how it compiles information from lots of articles so it doesn’t take forever to read.” 

But when asked whether they considered AI’s impact on the planet, most admitted they didn’t. Still, AI’s environmental footprint is far from insignificant. 

Large language models (LLMs), the type of AI behind popular chatbots like ChatGPT and Gemini, are “trained” on massive datasets, requiring enormous computing power. Once deployed, they continue to use significant resources every time they generate a response. According to the Harvard Business Review, training a single LLM can consume thousands of megawatt hours of electricity and emit hundreds of tons of carbon. 

And AI doesn’t just require electricity. It also demands fresh water and rare minerals. Data centers, which house the hardware powering AI, use millions of gallons of water for cooling systems which are fast to overheat. Meanwhile, the chips that power these systems rely on materials like cobalt and lithium, which are mined in a process with a host of its own environmental concerns. 

“The issue is that when data centers evaporate so much water so quickly, it throws off the cycle. More of it ends up in the atmosphere instead of where we need it on the ground, and that affects food, water, and energy,” Carrollton debate student Carlota Giampetruzzi ‘27 said. She argued that if the pattern continues, local resources could become strained. “The water these centers consume isn’t just for cooling. It’s water that could have gone towards other things. If we keep diverting it like this, we’re creating a chain reaction.” 

Despite this, many AI companies don’t disclose how much energy or water their systems consume. And if corporations aren’t talking about it, it’s no surprise that students rarely think about AI’s environmental footprint either.

At Carrollton, the conversation is beginning to shift as students start to consider AI’s hidden environmental costs, with debate students suggesting solutions. 

“Relocating data centers to colder climates could reduce water use for cooling,” said Lily Roos ‘27. “Alaska, for example, has naturally lower temperatures that could make operations less resource-intensive.” 

Roos emphasized the need to reframe how people discuss AI’s environmental toll. “Everyone talks about electricity, but water plays just as big a role,” she said. “If we relocated data centers to colder regions, we could ease some of that strain.” 

She also stressed that the problem goes beyond just water usage. “People act like the water just disappears, but it doesn’t. It goes somewhere.” Her point suggests that water is only one part of a larger cycle and that AI’s environmental footprint ripples through energy systems, agriculture, and even climate patterns. 

While students are starting to talk about AI’s footprint, policymakers are shaping its future. 

In July, President Trump signed three executive orders as part of a new AI Action Plan designed to accelerate AI deployment, loosen regulatory barriers, and promote U.S. “global AI dominance.” 

According to The Guardian, the plan emphasizes deregulation, including easing environmental reviews under the Clean Water Act and the National Environmental Policy Act (NEPA) in order to speed up construction of data centers. 

Meanwhile, Chris Wright, the new Energy Secretary, has expressed skepticism about climate change and rolled back climate-focused policies. 

Other countries are moving in a different direction. 

The European Union’s AI Act, passed in 2024, includes transparency requirements that will eventually force AI companies to disclose technical details such as energy consumption and data use. 

In Japan, the government has begun planning new standards for data centers, including proposals that facilities built after 2029 must meet energy efficiency rules or face penalties. And some companies are experimenting with hydrogen-powered or offshore “floating” data centers to reduce emissions. 

Still, AI’s environmental impact isn’t entirely negative. Advocates point out that AI can be used to model climate data, improve renewable energy efficiency, and even optimize water use in agriculture. Some argue that if applied responsibly, AI could become a tool in fighting the very climate problems it risks worsening. 

Cary Coglianese, a professor of Law and Political Science at the University of Pennsylvania, expressed optimism for AI’s future, citing it as a possible tool for solving societal issues. “What gives me hope for AI, then, is its potentially greater capacity to help solve complex policy problems that are becoming exceedingly difficult. Technology won’t be the only answer, but it is important to improve the information basis for decision making.” 

One thing is for sure: AI is here to stay. The question is whether we can balance its benefits with the toll on our planet.