It's a fair question, and a sign you care. The answer isn't a simple yes or no, and much of what circulates online is either exaggerated or brushed aside. Here's what reliable sources say, in plain English.
The short answer
AI does have a real and growing environmental cost, mostly from the electricity and water used by data centres and the emissions tied to building them. Your individual use of a chatbot is small, though. The bigger questions are about how fast data centres are growing, where their power comes from, and how they affect local communities.
How AI uses energy and water
AI runs on powerful computer chips in large buildings called data centres. Energy goes into two main jobs:
- Training: building a model, which takes a lot of computing over weeks or months.
- Inference: answering your questions, every time anyone uses it. With billions of requests, this adds up.
Data centres also need cooling, which can use water, and the buildings, chips and power plants all have emissions from manufacturing and construction.
The big picture: data centres
The International Energy Agency (IEA) is one of the most cited sources here:
- Data centres used about 415 terawatt-hours (TWh) of electricity in 2024, roughly 1.5% of global electricity (IEA, Energy and AI).
- In its Base Case, that more than doubles to about 945 TWh by 2030, slightly more than Japan's total electricity use today, with AI as the biggest driver (IEA).
- An updated IEA analysis projects data centre use roughly doubling from 485 TWh in 2025 to 950 TWh in 2030, around 3% of global electricity (IEA, Key Questions on Energy and AI).
- In the U.S., data centres account for nearly half of electricity demand growth to 2030 in the IEA's Base Case (IEA).
Two kinds of perspective matter. Globally, data centres remain a small share of electricity, and they account for around 10% of global demand growth to 2030; things like industry, electric vehicles and air conditioning grow more (IEA). Locally, though, they're concentrated: nearly half of U.S. data centre capacity sits in five regional clusters (IEA), which can strain local grids and raise questions about who pays for new power lines.
Your own use: how much is one prompt?
Google published one of the first detailed measurements. It estimates the median Gemini text prompt (May 2025) used 0.24 watt-hours of energy and 0.26 millilitres of water, about five drops, which it compares to watching TV for under nine seconds (Elsworth et al., Google, 2025).
Some important caveats:
- That's a median. Google's paper notes the distribution is skewed: a small share of prompts, such as those with many tokens, use far more energy than the median (Google). It also covers text prompts only, not images or video.
- The figures are Google's own measurements, not an independent audit.
- It doesn't include the energy used to train the model.
- Other companies haven't all published comparable numbers.
So: a typical chat is a small footprint per person. The total matters because of scale.
Emissions from tech companies
Big tech companies have reported rising emissions as they build out AI:
- Google reported its emissions were up 48% compared with 2019 in its 2024 report, citing data centre energy and supply chains (Google 2024 Environmental Report).
- Microsoft reported emissions of +23.4% vs 2020 (FY2024, 2025 report), attributing growth largely to data centre construction and supply chains (Microsoft 2025 Environmental Sustainability Report).
On the supply side, the IEA expects renewables to meet about half of the growth in data centre electricity demand, with natural gas and nuclear also playing a role (IEA). Google also says efficiency work cut the energy per median Gemini prompt by 33 times over one year (Google).
Where AI could help
The IEA also explores how AI might help the energy system itself, such as improving grid management and efficiency (IEA, Energy and AI). These benefits are possible but not guaranteed, and they don't cancel out the costs automatically.
What you can do
You don't need to feel guilty for using a chatbot. If you'd like to use AI more lightly:
- Use the right tool for the job. A regular web search or a calculator is fine for simple facts.
- Skip unnecessary generation, like making dozens of images or videos you won't use.
- Choose lighter modes when you don't need deep "reasoning."
- Pay attention locally. If a data centre is proposed near you, public meetings are where questions about water, power and costs get answered.
Common questions
Is AI bad for the environment?
It has a real and growing footprint from electricity, water and construction, but individual chatbot use is small. The impact depends heavily on scale and on how the power is generated.
How much energy does one AI prompt use?
Google estimates its median Gemini text prompt used 0.24 Wh and 0.26 mL of water (May 2025). Complex tasks use more.
How much electricity do data centres use?
About 415 TWh in 2024, around 1.5% of global electricity, projected to roughly double by 2030 (IEA).
Does AI use a lot of water?
Data centres use water for cooling. Per prompt it's small; locally, large data centres can matter, especially in dry regions.
Should I stop using AI?
That's your choice. Using it thoughtfully, for tasks where it actually helps, is a reasonable middle ground.