Carbon ShredderGet started

Is ChatGPT bad for the environment?

One question is small: about a third of a watt-hour, by OpenAI's account. The less comfortable parts are that OpenAI hasn't shown how it measured that, and that generated images and video cost far more per item than text.

Energy per querystated by OpenAI
0.34 Wh
Water per querystated by OpenAI
0.32 mL
CO₂e per queryour calculation
32 – 120 mg
Queries to match one burgerquarter-pound beef patty
82,000 – 300,000
Calculate yours

Your ChatGPT use, over a year

Text prompts
per day
Images generated
per week
Video clips generated
per month
Your AI use, per year

230 – 860 gCO₂e

Less than a tenth of a quarter-pound beef burger.

  • Text230 – 860 g
  • Images
  • Video

Water from text prompts: 2.3 L a year. Nobody publishes water figures for images or video.

ChatGPT: disclosed by the maker. An average ChatGPT query, as stated by OpenAI's CEO. OpenAI has published no methodology, model or prompt length behind it, and no emissions figure — the CO₂e range is our calculation from the energy figure.

For most people that is a rounding error. A typical footprint is measured in tonnes, and diet, driving, flying and home energy are where the tonnes come from.

What OpenAI has published

One sentence, no method

An average ChatGPT query, as stated by OpenAI's CEO. OpenAI has published no methodology, model or prompt length behind it, and no emissions figure — the CO₂e range is our calculation from the energy figure.

Disclosed by the maker

Energy per query
Yes
Emissions per query
No
Water per query
Yes
Methodology
No

ChatGPT next to other models

ModelSourceEnergyCO₂eWater
GeminiGoogleDisclosed by the maker0.24 Wh30 – 83 mg0.26 mL
ChatGPTOpenAIDisclosed by the maker0.34 Wh32 – 120 mg0.32 mL
Mistral Vibe (Mistral Large 2)Mistral AIDisclosed by the maker1.1 g45 mL
CopilotMicrosoftDisclosed by the maker0.16 – 0.6 Wh15 – 210 mg0 – 0.067 mL
ClaudeAnthropicNot disclosed0.16 – 0.6 Wh15 – 210 mg
Grok, Meta AI, DeepSeek & othersVariousNot disclosed0.16 – 0.6 Wh15 – 210 mg

See the full ranking and transparency scorecard

Images and video

Where a ChatGPT session gets heavy

OpenAI publishes no figures for image generation or its Sora video model. These are independent measurements of comparable open models.

One image

0.63 – 2.9 Wh

1.9 – 8.5 text queries

One short clip

30 – 940 Wh

89 – 2,800 text queries

Common questions

How much energy does a ChatGPT query use?
About 0.34 watt-hours, according to OpenAI's CEO — roughly what an LED bulb uses in a couple of minutes. OpenAI has not published the model, prompt length or methodology behind that figure.
How much water does ChatGPT use?
About 0.32 millilitres per average query, by OpenAI's own statement (0.000085 gallons) — a few drops. That figure, too, comes without a published methodology.
How much CO₂ does ChatGPT produce per question?
OpenAI publishes no emissions figure. Converting its 0.34 Wh with the grid range Google reports for its own data centres gives roughly 32 to 117 milligrams of CO₂e per query — our calculation, not OpenAI's.
Does ChatGPT image generation use more energy than text?
Much more per item. OpenAI publishes nothing for images, but independent measurements of image models put one image at roughly 0.6 to 3 watt-hours — several times a text query, or more.
Is ChatGPT worse for the environment than other AI?
It can't be settled with what is published. Google reports a lower figure for Gemini, but it is backed by a full methodology while OpenAI's is a single statement. Anthropic, xAI, Meta and DeepSeek publish nothing to compare against.
Can I offset my ChatGPT use?
You can fund verified carbon credits equal to it — or well beyond it, with an AI Climate Pass. It doesn't make your use "carbon neutral" — the emissions still happened — and for most people the amount is small next to diet and travel.

Sources

  1. Sam Altman (OpenAI) — The Gentle Singularity (2025-06-10)
  2. Google — technical paper on the energy, emissions and water of Gemini prompts (arXiv 2508.15734) (2025-08-21)
  3. MIT Technology Review — investigation into the energy use of AI text, image and video models (2025-05-20)
  4. Luccioni, Jernite & Strubell — Power Hungry Processing (FAccT 2024) (2024)

Last reviewed 2026-09-15. Carbon Shredder is not affiliated with OpenAI.