Is ChatGPT Bad for the Planet?

Hot take: You might be using more energy doom-posting about AI than ChatGPT used to help you write it. Let’s break down the real energy and water use, some surprising upsides, and what we can realistically do to consume less.

Water Use: Thirst Traps
Each session with ChatGPT, you’re using an estimated 16 oz or one bottle of water. It’s needed to cool down data centers that store the high-powered servers needed to power AI. Fresh water is used because salt water is corrosive, and it immediately turns into steam through a process called evaporative cooling, leaving little that can be captured and reused.

Has AI increased water useI? Yes, but not as much as you may think. Journalists looked into Microsoft’s water use in West Des Moines, Iowa, where one of the major data centers powering ChatGPT is located, and found that it accounted for about 2% of the city’s total water consumption. Microsoft pays the same rates as residents and also invests in local irrigation and infrastructure improvements.

Energy Use: AI’s Power Bill
In the US, AI has increased energy use about 4% projected to be 12% by 2028. The International Energy Agency forecasts that globally, AI will only create a 10% overall increase by 2030. But it’s all over the news here in the U.S. because those numbers are projected to jump as high as 50% by 2030, far outpacing the rest of the world’s AI energy use. Richmond, Virginia, for example, is the world’s data center capital, home to more than 300 centers. For reference, Beijing is a distant second. Richmond is where the internet was born and sits close to federal agencies. Those centers consume 20% of the city’s power but deliver 30% of its tax revenue.

AI may actually be saving energy. Think about it like this: a human completing a task that an AI can do in seconds uses 30-80 times more emissions. You’re sitting at your computer all day, with all your devices, spending 8 hours outlining and drafting an article AI could have written in a fraction of that time.

Renewable Energy: Unreliable
Can’t we just use renewable solar and wind energy? Nope. AI needs energy 24/7 and wind and sun are never constant. A silver lining is that geographically speaking, that power source can come from anywhere. Data centers don’t have to be local. Unlike driving a car, which releases emissions wherever you physically are, AI’s energy footprint can be moved. Relocated to regions with cleaner grids, placed underwater for natural cooling, or even powered by nuclear facilities. But for now we’re all still just girls living in a fossil fueled world. AI doesn’t just use energy; it helps find more of it. That means more drilling and a longer dependence on fossil fuels.

Efficiency: Overachiever Mode
Now for the good news: AI can improve grid efficiency, optimize cooling, monitor air and water quality, and strengthen recycling and waste systems—easing the burden and improving compliance. Another boon to the environment is game-changing monitoring of endangered species, biodiversity, migration patterns, and poacher activity. We’ve never had a way to process such massive streams of data from sensors, drones, and satellites, so AI finally lets us see patterns (and threats) in real time.

Resources: Machines for Mother Nature
Materials discovery is another area where AI excels—identifying new mining opportunities and helping protect remote sites from illegal activity. It’s also being used to assist in predicting natural disasters plus response and recovery through damage assessment, route optimization, and coordination. The city of LA, where I live, even launched a fire chatbot after the Palisades fires…and it’s not good. As one news headline put it, “California’s fire protection agency made an AI chatbot. It can’t answer one crucial question.” So all of these promises come with a grain of salt and as copywriters we know the devil is in the details and execution is everything.

Innovation: We’ve Seen This Show Before
Tech bros like to remind us we all said the same thing about the internet boom: every time you ordered a book online from Amazon, you were burning a lump of coal. And when social media and online gaming became big, think 24/7 streaming on Twitch, the need for more power only served to increase innovation.

Retail’s Role: Shop ’Til the Planet Drops
Whether it’s more shopping, more advertisements, more scrolling and screen time, AI will increase consumerism. Do retailers understand the invisible environmental cost? I doubt the industry that brought us fast fashion, influencer hauls, and greenwashing are going to be the ones leading that conversation.

Regulations: No Rules, Just Vibes
Ireland is one of the only nations with an AI environmental policy stating that any new grids need to bring the same amount of energy to the grid as it takes. Meanwhile, the EU is focused on transparency and reporting, requiring large AI systems to disclose energy use, environmental impact, and training resources. It’s not a cap or a quota, but it forces companies to show their math, which is the first step toward accountability. In the U.S., there is no federal environmental framework for AI yet.

Transparency: Don’t Ask, Demand
It helps new technologies build on work that has already been done and energy that has already been consumed. There are more efficient models than ChatGPT like DeepSeek but you pay in creativity and reliability. DeepSeek made headlines by claiming to use a fraction of the power of ChatGPT but it was also built on the foundation of an existing model, Meta’s LLaMa.

Your Turn: What Can You Do?

Ask yourself, is this a job for AI or something else? Asking ChatGPT uses 10x the amount of energy as Googling it, because the models have billions of parameters at work.
Do you really need a super-computer to help write flash sale subject lines? (I am 100% guilty of this.)
Smarter prompting. Train one channel or create one Project for every topic, project, client, etc. That way you don’t have to reset the scene every time.
If you mess up a prompt, you can immediately hit stop on the text generation then hover over your f-ed up prompt and click on the pencil button to edit and resubmit.
If it’s taking a long time to “think” on something relatively basic, hit Skip.

We’re not powerless bystanders in this story. We’re the people deciding what gets written, sold, scaled, advertised, or abandoned altogether.

SOURCES:

Laura Cozzi: IEA’s Energy AI Report
https://podcasts.apple.com/us/podcast/ai-energy-and-climate-podcast/id1798264288?i=1000703490870
How AI affects our environment
https://www.youtube.com/watch?v=ZmRT7e25Kxg

The growing environmental impact of AI data centers’ energy demands

https://www.youtube.com/watch?v=-sNKfRq1oKg&utm_source=chatgpt.com

Peter Henderson: Environmental Impact of AI (and What Developers Can Do)

https://www.youtube.com/watch?v=q78KYBps-zY&utm_source=chatgpt.com

Climate Impact of Generative AI

https://www.youtube.com/watch?v=FLUyLqGlW2s&utm_source=chatgpt.com

Does Compute | Environmental Impact of AI

https://www.youtube.com/watch?v=lv4etQ3xZpY&utm_source=chatgpt.com

The Machine Ethics podcast – DeepDive: AI and the environment