‘People are often Curious about how much energy a chatgpt inquiry uses, “Openai CEO Sam Altman wrote in a long blog post last week. The average query, Altman wrote, used 0.34 watt-hour energy:” which would use an oven within a little more than one second, or use a high-effective bulb in a few minutes. “

For an enterprise with 800 million active users (and growth) weekly, the question becomes how much energy all these searches use, more and more urgent. But experts believe that the figure of Altman does not mean much without much more public context of Openai on how it came to this calculation – including the definition of what an ‘average’ query is, whether it includes image generation, and whether Altman includes additional energy consumption, such as from the training of AI models and the cooling of Openai’s servers.

As a result, Sasha Luccioni, the climate leader at AI Company Hugching Face, places not too much stock in Altman’s number. “He could have pulled it out of his hole,” she says. (Openai did not respond to a request for more information on how it came to this number.)

While AI is taking over our lives, it is also promising to transform our energy systems, facing the carbon emissions while trying to fight climate change. Now a new and growing research is trying to put hard numbers on how much carbon we actually radiate with all our AI use.

This effort is complicated by the fact that big players like Openai disclose little environmental information. An analysis presented by Luccioni for peer review this week, and three other writers look at the need for more environmental transparency in AI models. In the new analysis of Luccioni, she and her colleagues use data from OpenRouter, a ranking of a major language model (LLM) traffic, to determine that 84 percent of LLM use in May 2025 was for models with no disclosure of the environment. This means that consumers choose overwhelming models with completely unknown environmental impacts.

“It stops me to buy a car and know how many miles per gallon it consumes, but we still use all these AI instruments every day and we have absolutely no effectiveness statistics, emission factors, nothing,” Luccioni says. “It’s not mandate, it’s not regulatory. Given where we are with the climate crisis, it must be the pinnacle of the agenda for regulators everywhere. ‘

Because of this lack of transparency, Luccioni says, the public is exposed to estimates that make no sense, but which are considered gospel. For example, you may have heard that the average chatgpt request takes ten times as much energy as the average Google search. Luccioni and her colleagues keep this claim to a public remark that Alphabet’s chairman John Hennessy was made in 2023 the parent company of Google.

A claim made by a one -company management member (Google) on the product of another company he has no connection (Openai) is at best – she finds Luccioni’s analysis, and this figure is repeated in the press and policy reports. (While I wrote this piece, I got a pitch with these exact statistics.)