Every Time You Ask ChatGPT Something, It Uses As Much Electricity As An Oven Does In One Second, Says Sam Altman: The 'Cost Of Intelligence' Will Only Lower

OpenAI chief executive Sam Altman says that a ChatGPT query consumes roughly 0.34 watt-hours of electricity, which is “about what an oven would use in a little over one second."

What Happened: Altman, in a blog post he published on Tuesday, also mentioned that each ChatGPT query consumes "about 0.000085 gallons of water,” which is roughly one-fifteenth of a teaspoon.

Altman argues that "the cost of intelligence should eventually converge to near the cost of electricity," positioning the chatbot as comparatively frugal even as environmental watchdogs scrutinize AI's hidden climate bill.

OpenAI has not publicly shared details of their methodology, but the disclosure lands as researchers warn that data-center power demand could outstrip Bitcoin Mining by next year. One study from Vrije Universiteit Amsterdam projects AI could soon swallow nearly half the electricity flowing into global server farms.

See also: Elon Musk Says He ‘Regrets’ War Of Words With Trump: ‘Went Too Far’

Water use complicates the equation. A Washington Post investigation found that drafting a 100-word email with GPT-4 required "a little more than one bottle" of water, with consumption varying widely by data-center location. MIT researchers likewise warned in January that generative models are driving up both energy and water footprints as companies race to scale.

Why It Matters: The Department of Energy projects that U.S. data centers may consume up to 12% of the nation’s electricity by 2028, a share that could climb as AI adoption accelerates.

Altman's efficiency pitch follows corporate pledges to curb resource demand, yet Yale's Environment 360 notes hyperscale facilities still guzzle millions of gallons annually. Planet Detroit adds that generative systems may use 33 times more energy than conventional software for the same task.

Whether the teaspoon math holds up, Altman insists smarter algorithms and automated chip fabrication will keep driving the price and resource cost of "intelligence" lower.

Image via Shutterstock

Read next: Anduril CEO Palmer Luckey Confirms IPO Plans To Compete For Trillion-Dollar Defense Contracts: ‘We Are Definitely Going To Be A Publicly Traded…’