The Hidden Cost of Artificial Intelligence

It’s 11:30 pm on a school night, and a senior has calculus homework due the next morning that they just can’t solve. Their eyelids begin to drift shut, but there are still five more questions sitting blankly on the paper. Then, suddenly, a flash of brilliance. They open a ChatGPT tab, type the equation into the chat box, and at long last, solve the problem. The impossible homework was solved in under a minute.

What the student couldn’t see was what happened behind the screen. AI servers processed the prompt and spit out an answer in seconds, but the servers also heat up and require water to cool them.

Each prompt an AI model receives requires its servers to use electricity, which then incites the need for cooling. The amount per prompt is small at first, but adds up quickly. According to Business Energy UK, ChatGPT uses an estimated 39 million gallons of water per day. This is equivalent to the amount of water used if every person in Taiwan flushed their toilet at once.

When AI servers use water, it is often not reusable because it can become contaminated with minerals or chemicals during the cooling process.

Because the water use occurs behind the scenes, it can be difficult to visualize what happens to it. Miles Stagman ‘26, former head of Latin’s Green Club, said, “People can understand energy loss a lot easier, but it feels like water can just always be recycled.”

Miles’s observation leads to a larger reason for AI’s water use going unnoticed. One reason the issue is ignored is that people comprehend energy waste far better than water waste. We hear about carbon footprints all the time, like in science class, or when we hear about how many flights celebrities take, but we rarely hear about water footprints.

Because water feels renewable, it’s easy to assume that the water used to cool servers simply comes back into the ecosystem and can be used again.

Latin has a comprehensive AI policy, but it is mainly focused on students cheating in school.

The challenge for our community is more than just deciding when AI is being used as a tool or as plagiarism; it’s recognizing the consequences of its convenience.

“If I’m confused about math or any other class, I usually won’t use AI because I feel like it hinders my ability to learn,” senior Ehren Curry said. “Instead, I’ll go to the Math Center, but I do that more for my learning than thinking about water usage.”

While some students, like Ehren, think of their learning comprehension first, the increasing capabilities of AI make it more integrated into our lives and harder to stop using. Using Ehren’s example, AI can make it really easy for a student to relearn a math concept when they don’t have access to a teacher.

Solutions to the overuse of AI for simple tasks start small. Upper School biology teacher David Wisnieski suggested to his biology students that they turn off the AI overview that Google automatically shows when you search.

“All searches use energy, of course, but AI uses more energy, and Google adds it automatically now,” Mr. Wisnieski said. “You can add ‘-AI’ to a search query to exclude AI. You can also use search engines like DuckDuckGo that don’t use AI unless you opt in.

While tips like these may seem insignificant, each prompt has an impact, so every small change can, too. Even if we take away one AI usage per day at a time, it adds up, just like the amount of prompts we use adds up to the overall water usage. There is more to the conversation around AI than cheating.

“People talk about AI as only an academic ethics issue,” Miles said, “but it’s also a huge environmental issue.”