Is AI bad for the environment?
Yes, it has a real footprint. No, your one ChatGPT question isn't the villain.
You're right that data centers use a lot of energy and water. That's not propaganda — it's measurable. A medium-sized data center can use about 110 million gallons of water a year, roughly what 1,000 households use. US data-center electricity demand is projected to nearly double from 2025 to 2028.
Here's what's more useful than guilt: context. A single ChatGPT query uses roughly the energy of running an LED bulb for a few minutes. Streaming a 4K movie for an hour, a 10-minute hot shower, or driving to class uses dramatically more. The environmental question isn't really "should I use AI?" — it's "should society build 150 gigawatts of new data centers, and how should they be powered and cooled?" That's a policy and infrastructure question, not a personal-virtue question.
The real leverage isn't skipping a prompt. It's caring about where data centers get built, what energy source they use, and whether communities near them get clean water. Asking those questions publicly is more impactful than any personal opt-out.
Every AI question boils a bottle of water.
One question ≈ a few seconds of AC. The scale problem is real; the individual prompt isn't.
- Batch your questions. Ten focused prompts use less than forty lazy ones.
- Look up where your favorite AI's data centers are. Some run on renewables; some don't.
- Read local news about data center permits in your town — that's where actual decisions happen.
- Ask Otis: "Which AI company has the cleanest energy footprint right now?"
Still sitting with something?
Otis can get specific about your situation, your field, your campus.