How Much Energy and Water Does ChatGPT Use? (2026 Facts)
How much energy does ChatGPT use? OpenAI's chief executive Sam Altman says the average ChatGPT query uses about 0.34 watt-hours of electricity and roughly one fifteenth of a teaspoon of water. Independent estimates from Epoch AI and Google's figures for Gemini land in a similar range, although none of these numbers comes from an independent audit.
If you have wondered how much energy ChatGPT uses, the short answer is that one question takes very little, while the total across roughly 1 billion weekly users, a figure Epoch AI gives for June 2026, is large and growing. This guide, checked on 22 October 2026 against Sam Altman's own figures, Epoch AI, Google and the International Energy Agency, sets out the numbers, explains why they differ and says what nobody can yet prove.
In this guide
How much energy does one ChatGPT query use?
The only per-query figure from OpenAI's leadership is Sam Altman's. In a blog post dated 10 June 2025, he wrote that the average query uses about 0.34 watt-hours, about what an oven would use in a little over one second or a high-efficiency light bulb in a couple of minutes. The post gives a single average and does not explain how it was worked out, and we did not find a newer official figure when we checked.
A watt-hour is the energy a 1 watt device uses in an hour, so 0.34 watt-hours is tiny on its own. By simple arithmetic, 100 queries at that average would use 34 watt-hours.
Epoch AI, an independent research group, estimated in a February 2025 analysis that a typical GPT-4o query uses roughly 0.3 watt-hours, about ten times less than an older, widely repeated estimate. Its July 2026 overview of AI energy use puts a 100-word prompt to a free chatbot at around 0.6 watt-hours once a share of the energy used to develop the model is included.
Source | What it says | Date | Caveat |
Sam Altman (OpenAI) | About 0.34 Wh for the average ChatGPT query | 10 Jun 2025 | One average, method not given |
Epoch AI | About 0.3 Wh for a typical GPT-4o query | Feb 2025 | Independent estimate |
Epoch AI | About 0.6 Wh for a 100-word prompt to a free chatbot, including a share of model development | Jul 2026 | Chatbots in general, not ChatGPT alone |
0.24 Wh for a median Gemini text prompt | Aug 2025 | Gemini, not ChatGPT, and not independently verified |
How much water does ChatGPT use?
Altman's post gives the same kind of average for water: about 0.000085 gallons per query, roughly one fifteenth of a teaspoon, or around 0.32 millilitres. Data centres often use water for cooling. Epoch AI explains that densely packed chips give off intense heat, so in modern data centres a liquid coolant carries the heat away and passes it to a stream of water, which releases it outdoors.
Google's published figure for a median Gemini text prompt is 0.26 millilitres of water, about five drops, in its August 2025 methodology post. The two numbers are close, which suggests the order of magnitude is right. Each company is describing its own systems, though, and we found no independent check of either.
Why do the estimates differ?
Epoch AI says the energy for a chatbot reply depends on the length of your prompt, which model answers, how much it thinks first and how long the answer is. A short question to a free chatbot sits at the low end. A long document, a model reasoning for minutes or an AI agent doing a multi-step job sits much higher, and Epoch notes that coding agents use substantially more energy than chatbots.
Method matters as well. These are the main reasons two honest sources can give different numbers:
What is counted. Google found that counting only the active chips gives 0.10 Wh for a median Gemini prompt, while adding idle machines, host computers and data centre overhead gives 0.24 Wh.
Training. Most per-query figures leave out the energy used to train the model. Epoch's 0.6 Wh figure includes a share of it.
Averages hide the spread. An average or median says nothing about the heaviest requests.
Age. Altman's figure is from June 2025, and ChatGPT's models have changed since. We did not find an updated official number.
How does it compare with everyday things?
Putting the numbers next to familiar objects helps, as long as you remember they are estimates:
Altman compares the average ChatGPT query with an oven running for a little over one second.
Epoch AI says 0.6 Wh is less energy than a microwave uses in 10 seconds, or a 10 watt light bulb uses in five minutes.
Google compares its 0.24 Wh median Gemini prompt with watching television for under nine seconds.
Epoch AI calculates that an average American household would need to send over 400 chatbot messages a day to add 1% to its daily energy use of around 29,000 Wh. That is a US figure.
For a wider view of how people use the tool, see our ChatGPT statistics for 2026, and for how a rival's numbers look, our ChatGPT vs Gemini comparison.
What about the bigger picture of data centres?
One query is small. The total is not. The International Energy Agency says data centres used around 1.5% of the world's electricity in 2024, or 415 terawatt-hours, with the United States taking 45%, China 25% and Europe 15%. It expects that to more than double to around 945 TWh by 2030, slightly more than Japan's total electricity use today, and to reach around 1,200 TWh by 2035.
The IEA also puts this in proportion. Data centres account for around one-tenth of global electricity demand growth to 2030, less than industrial motors, air conditioning or electric vehicles. A typical AI-focused data centre consumes as much electricity as 100,000 households, and the largest ones now being built will use 20 times as much.
On emissions, Epoch AI cites a 2025 IEA report saying data centres account for less than 1% of global CO2 emissions, with AI only part of that. It also warns that local effects can be real even when the global share is small. As examples, it points to nitrogen dioxide levels that rose near a gas-powered data centre in Memphis, and to disagreement among analysts over how far data centres are raising electricity prices.
These are mostly global and US figures. We did not find a published UK-specific number for ChatGPT queries, so treat any claim that gives one with caution. For more on how the technology works, read how ChatGPT works and our plain-English guide to large language models.
Can you use ChatGPT with less energy?
Your own footprint from chatting is small, so this is not a reason to stop using it. If you would still like to be thoughtful, these habits follow from the factors Epoch AI lists. They are our suggestions, not OpenAI advice.
Ask focused questions, and paste the part of a document you need rather than the whole thing.
Use a lighter, faster model for simple jobs and keep longer thinking modes for problems that need them.
Ask for the length of answer you actually want.
Avoid regenerating answers out of habit.
If you are also weighing up whether ChatGPT is safe to use with personal information, our privacy guide covers that, and our comparison of the best AI chatbots shows the main alternatives.
Energy and water FAQs
How much energy does ChatGPT use per query?
Sam Altman wrote in June 2025 that the average ChatGPT query uses about 0.34 watt-hours. Epoch AI's independent estimate for a typical GPT-4o query was roughly 0.3 watt-hours. Longer prompts, longer answers and reasoning modes use more.
How much water does a ChatGPT question use?
Altman's figure is about 0.000085 gallons per query, which is roughly one fifteenth of a teaspoon or 0.32 millilitres. Google reports 0.26 millilitres for a median Gemini text prompt. Neither number has been independently verified.
Is ChatGPT's energy use bad for the climate?
One query is tiny, but the total is growing. The International Energy Agency says data centres used about 1.5% of the world's electricity in 2024 and could reach around 945 TWh by 2030. Epoch AI cites the IEA in saying data centres produce less than 1% of global CO2 emissions today.
Does training ChatGPT use more energy than answering questions?
Training a large model is particularly energy-intensive, according to Epoch AI, but the energy use does not stop there because every answer needs computing power too. We did not find an official figure for the training energy of the models behind ChatGPT.
Can I trust OpenAI's energy and water figures?
Treat them as an order of magnitude. Altman gives a single average without explaining the method, and Google says its own Gemini figures have not been independently verified. The independent Epoch AI estimates sit in a similar range, which is reassuring but not proof.
Is there a UK figure for ChatGPT's energy use?
We did not find a published UK-specific figure for ChatGPT queries. The numbers in this guide are global averages or come from US-focused analysis, so the real figure for a UK query depends on where the servers are and how they are powered.
Key takeaways
OpenAI's CEO puts the average ChatGPT query at about 0.34 Wh and one fifteenth of a teaspoon of water (June 2025).
Independent estimates from Epoch AI (0.3 to 0.6 Wh depending on what is counted) and Google's Gemini figure (0.24 Wh) are in a similar range.
Longer prompts, longer answers, reasoning modes and AI agents use more than a quick question.
The IEA expects data centre electricity use to more than double to around 945 TWh by 2030, from 415 TWh in 2024.
No source we found has independently audited OpenAI's figures, and we found no UK-specific number.
The bottom line
One ChatGPT question uses a very small amount of energy and water, on the best figures available, but the combined demand from AI is real and rising fast. Expect better data as companies are pressed to publish measured, comparable numbers. Explore more ChatGPT guides on whatischatgpt.co.uk, starting with what ChatGPT is and whether it is free.
