AI Carbon & Water Footprint Calculator

Turn your daily prompts into a yearly energy, carbon and water estimate.

Every AI prompt runs on a datacenter that draws power and uses water to stay cool, so a habit of chatting with a model all day has a footprint. This tool gives a rough estimate of that footprint from four things: how many prompts you send a day, how long the replies are, the kind of model behind them, and how many people you are counting. It answers questions like how much energy does ChatGPT use, what the AI water footprint looks like over a year, and the carbon footprint of an AI query in terms you can picture. The figures are deliberately rough — providers rarely publish exact per-prompt numbers and independent studies disagree — so read it as an order-of-magnitude guide, not a meter reading.

Read the guide: How Much Energy Does ChatGPT Use Per Prompt?

Your AI usage

Response length
Longer replies generate more tokens, so they use more energy.
Model class
Reasoning models think out loud and draw roughly ten times a large model per answer.

Estimated carbon per year

4.38 kg

19.71 L of water · 10.95 kWh of energy

Energy / day
30 Wh
Water / day
54 mL
CO2 / year
4.38 kg
Water / year
19.71 L

A year of that is roughly

  • Google searches36,500
  • Phone charges730
  • Miles driven10.95
  • Water bottles (500 mL)39

These are rough estimates. Providers rarely disclose exact per-prompt energy, and independent figures vary widely — the numbers use mid-range values (small ~0.3 Wh, large ~1.5 Wh, reasoning ~15 Wh per medium reply; 0.4 g CO2 and 1.8 mL water per Wh). Treat them as order-of-magnitude, not measured.

How it works

  1. 1

    Set your prompt volume

    Enter how many prompts you send in a day and how many people to count. One person exploring ideas looks very different from a 500-seat company, and the yearly totals scale with both.

  2. 2

    Pick reply length and model class

    Choose short, medium or long replies and whether you use a small, large or reasoning model. Longer replies generate more tokens, and reasoning models draw roughly ten times a large model per answer, so both push the energy figure up.

  3. 3

    Read the footprint and comparisons

    The headline is the estimated carbon per year, with water and energy alongside. Below it, a year of that usage is translated into Google searches, phone charges, miles driven and bottles of water so the numbers mean something.

Instant & 100% private — nothing is uploaded

Every calculation runs locally in your browser. The prompts, token counts and numbers you enter stay on your own device and are never sent to a server — nothing is stored, logged or shared.

Frequently asked questions

How much energy does one ChatGPT prompt use?
Common estimates put a short query around 0.3 Wh and a longer reply on a big model at a few watt-hours. Reasoning models that generate a lot of hidden tokens can use far more. This tool uses mid-range figures of about 0.3 Wh for a small model, 1.5 Wh for a large one and 15 Wh for a reasoning model at a medium reply length.
How is the water footprint worked out?
Datacenters use water to cool their servers. A common figure is about 1.8 litres per kilowatt-hour of energy, so the tool multiplies your estimated energy by that rate. It covers cooling water only, not the water used to generate the electricity itself, so the true figure is likely higher.
Why are the numbers only estimates?
Providers almost never publish exact energy or water use per prompt, and independent estimates vary by an order of magnitude depending on the model, the hardware and the grid. The constants here are reasonable mid-range picks, so the result is a ballpark for comparison and awareness, not a precise measurement of your usage.
How do the comparisons help?
Raw watt-hours and grams of CO2 are hard to picture. Converting a year of prompting into Google searches, phone charges, miles driven and water bottles gives a sense of scale. They are approximations too — a Google search is roughly 0.3 Wh, a phone charge about 15 Wh, a mile of driving around 400 g of CO2 — so use them to grasp magnitude, not to audit anyone.

Important

For planning and estimates only. Prices come from a published rate table dated on the page; providers change pricing without notice, and token counts here are approximations. Confirm against the provider’s own pricing before you budget or commit.

More tools

AI Token CounterPaste text and see an estimated token count, characters, words and input cost per model, live as you type.LLM Cost CalculatorPrice a single LLM API call or a whole batch: pick a model, set input and output tokens, get the cost.Meeting Cost CalculatorPut a dollar figure on any meeting from attendees, average pay and duration — with a live ticking burn meter.AI Chatbot Cost CalculatorWork out a chatbot's monthly and yearly API bill from daily messages, tokens per message and the model.AI GPU Cost CalculatorEstimate cloud GPU cost from wall-clock hours, GPU count and the price per GPU-hour.AI Model ComparisonSort and filter LLMs by input price, output price and context window in one live table.

More from the Hivly network

Free sister tools on our other sites.

Length ConverterMetres, feet, miles, inches, km and more, instantly.Scientific CalculatorPowers, roots, trig, logs, memory and history.Age CalculatorExact age in years, months and days, plus your next birthday.