Short on time? Here’s the whole article in four lines.
- One text question is small. By Google’s estimate, it uses about as much energy as watching TV for nine seconds.
- The total is growing fast, though it’s still far smaller than the impact of flying or food and farming.
- What you make matters. Video and lots of images use far more energy than text.
- You can use AI and still care about the planet. Use it on purpose, choose the right tool, and skip waste.
AI runs in data centers, which are huge buildings full of computers. Those computers need electricity to run and, often, water to stay cool.
| Concern | What’s going on |
|---|---|
| Electricity | Data centers used about 1.5% of the world’s electricity in 2024. The International Energy Agency expects that to more than double by 2030, to about what Japan uses today. |
| Water | Many data centers use water for cooling, and more is used by the power plants that supply them. Many new ones are being built in drought-prone areas. |
| Carbon | Emissions depend on where the power comes from. Coal and gas mean more carbon. Solar, wind, and nuclear mean less. |
| Local communities | Neighbors worry about water supplies, power bills, and noise. In the U.S., some towns have paused new data centers. |
| Hardware | Making chips and servers takes energy and materials, and old equipment becomes e-waste. |
The honest answer is that it depends, and companies don’t share much data.
| What you ask for | Rough energy use |
|---|---|
| A typical text question | Small. Google says a median Gemini text prompt uses about 0.24 watt-hours and about five drops of water. |
| A long or complex task | More. Bigger models, long documents, and “deep research” modes do much more work. |
| An image | More than a typical text answer. |
| A short video | Much more. One test found a 5-second AI video used about 700 times the energy of a single high-quality image. |
A grain of salt: Google’s figure is its own estimate. Researchers point out that it leaves out things like building the hardware and the heaviest requests. Treat every per-prompt number as a rough guide.
Here’s how data centers compare with other parts of modern life worldwide.
| What | Share of the world’s total |
|---|---|
| Data centers (all uses, not just AI) | About 1.5% of electricity and about 0.5% of energy-related carbon emissions |
| Air travel | About 2.5% of carbon emissions |
| Food and farming | About 26% of greenhouse gas emissions and 70% of the fresh water people use |
Data centers are a small slice today, but one of the fastest-growing. The International Energy Agency expects their emissions to roughly double as a share of the total by 2035. It also estimates that AI could cut more emissions in other industries, like energy and manufacturing, than data centers produce. That’s possible, not guaranteed.
Both of these are true:
So you don’t need to feel guilty about asking AI to help with an email. It’s still worth using it thoughtfully, and paying attention to how the industry grows.
| Do | Why it helps |
|---|---|
| Use AI on purpose, for tasks where it actually saves you time | Skips work that doesn’t need doing |
| Get it right in fewer tries. Give clear context up front. | Fewer redos means less computing. See How to Talk to AI: Prompting Basics. |
| Pick the right size tool. Use quick modes for simple questions. | Deep research and the biggest models use more |
| Go easy on video and bulk images | These are the heaviest tasks |
| Use a search engine for simple facts like store hours | A regular search is usually lighter |
| Look at a company’s commitments on clean energy, water, and transparency | Your choices and questions push the industry |
If sustainability matters to your customers, be open about how you use AI, just as you are about other choices. A line on your website about using AI thoughtfully goes further than staying silent or overpromising.
Avoid calling your business “carbon neutral” or “green” because of your AI choices unless you can back it up. Vague green claims can backfire.
Rosa runs a small natural soap company, and her customers care about the environment. She’d been generating dozens of AI product images a week, picking one, and tossing the rest.
She changed her routine. Now she writes one clear, detailed prompt, makes a few images, and uses her own phone photos for product shots. She uses AI mostly for text, like descriptions and emails, and she added a short note about her approach to her About page. Her customers appreciated the honesty.
Checked October 28, 2026.
