How Much Energy Does AI Use? The People Who Know Aren’t Saying

“People are often Curious to know how much energy uses a chatgpt query, ” Sam AltmanThe CEO of OPENAIwrote in one side side in a Long blog arte Last week. The average request, Altman wrote, uses 0.34 Watthers of energy: “About that an oven would use a little more than a second, or a high efficiency bulb would use in minutes.”

For a company with 800 million weekly active users (and growth), the question of the amount of energy that all of this research uses is becoming more and more pressing. But experts say that Altman’s figure does not mean much without much more public context of OpenAi on the way in which it has happened to this calculation, including the definition of what an “average” query is, whether or not it includes the generation of images, and whether or not Altman includes additional energy consumption, such as training models of training and servers of coolant.

Consequently, Sasha Luccioni, the head of the company AI Hugging Face, does not put too much stock in the Altman issue. “He could have removed this from his ass,” she said. (Openai did not respond to a request for more information on how she came to this number.)

While AI takes precedence over our lives, it is also promised to transform our energy systems, on the overeating of carbon emissions, while we are trying to fight against climate change. Now, a new number of growing research is trying to put a hard number on the quantity of carbon that we actually emit with all of our use of AI.

This effort is complicated by the fact that the main players like Openai disclose little environmental information. An analysis subject to peer exam this week by Luccioni and three other authors examine the need for greater environmental transparency in AI models. In the new analysis of Luccioni, she and her colleagues use data from OpenA classification of large -language model traffic (LLM), to note that 84% of the use of LLM in May 2025 was for models with zero environmental disclosure. This means that consumers massively choose models with completely unknown environmental impacts.

“It blows me that you can buy a car and know how many kilometers by Gallon it consumes, but we use all these tools of AI every day and we have absolutely no measure of efficiency, emission factors, nothing,” explains Luccioni. “It is not mandated, it is not regulatory. Given where we are in the climate crisis, it should be at the top of the agenda of regulators everywhere. ”

Due to this lack of transparency, says Luccioni, the public is exposed to estimates which have no meaning but which are considered as a gospel. You may have heard, for example, that average chatgpt demand takes 10 times more energy than research on average Google. Luccioni and his colleagues find this claim to a public remark that John Hennessy, the president of Alphabet, the mother company of Google, made in 2023.

A complaint made by a member of the board of directors of a company (Google) on the product of another company to which he has no relationship (OPENAI) is at best tenuous – but finds the analysis of Luccioni, this figure has been repeatedly repeated in press and policy reports. (While I was writing this piece, I had land with this exact statistic.)