What the Tech: AI environmental impact
AI may feel invisible. Just a few keystrokes and a quick reply from a chatbot. But behind the scenes, it is powered by enormous data centers running constantly.
Researchers say that training a single large language model, like ChatGPT, can use as much electricity as 100 U.S. homes use in an entire year. It sounds extreme, but it is not an exaggeration. The energy needed to process, train, and run AI models adds up quickly. And it is not just electricity. Data centers generate heat, and all that equipment has to stay cool. That requires water. And lots of it.
“When it is being trained, what is happening is it is a huge amount of data that is going in to make it what it is that you interact with,” says Keyona Meeks, an AI strategist. She helps organizations develop AI responsibly and track its real-world impact.
Even after training is complete, Meeks says AI still uses significant resources every time we interact with it. Whether you are planning a trip, writing a resume, or generating an image, “each interaction is like tiny sparks,” she says. “Multiply that by millions of users, and the environmental cost adds up.”
And most of that electricity still comes from fossil fuels.
A study from the University of California found that a single prompt to a chatbot like ChatGPT or Gemini can use about half a liter of fresh water just to cool the servers. Another estimate suggests the energy used per prompt is similar to running a lightbulb for a few minutes. That may not seem like much, until you consider how often these tools are used globally. According to OpenAI CEO Sam Altman, even saying “thank you” after every prompt can result in millions of dollars in additional energy costs each year.
Keyona Meeks adds, “AI’s environmental impact, as we are positioned right now, will be… I do not want to say cataclysmic, but it is not intellectually dishonest to say that word.”
Environmental groups are urging AI companies and data center operators to commit to renewable energy and more sustainable practices. While companies like Google and Microsoft have announced plans to reduce their footprint, the rapid growth of AI is pushing energy and water demands to new highs.
That pressure is already being felt here in Tennessee. In Memphis, Elon Musk’s AI company xAI is operating dozens of methane gas turbines to power a supercomputer in the Boxtown neighborhood. Residents and environmental groups have filed lawsuits, claiming the facility is releasing hazardous pollutants and placing additional health burdens on already vulnerable communities.
Meeks believes awareness is the first step toward responsible AI use. “We all have a responsibility to take a second to confirm the information that we are looking at,” she says. “Look for a second or third source that confirms it before you hit that share button.”
AI is changing how we work, create, and communicate. But as it gets smarter, it is worth asking whether it is getting any cleaner.
