ChatGPT uses energy similar to half of Denmark’s annual electricity consumption – Scandasia
Nordic News, Business and Life in Asia
The electricity consumed annually by ChatGPT to answer user prompts alone is now roughly equivalent to the total yearly electricity consumption of Iceland, or about half of Denmark’s annual electricity use, according to a new analysis by BestBrokers.
The findings come as OpenAI has recently introduced GPT-5.2, its most advanced artificial intelligence model to date. While the new model promises improvements in general intelligence, coding and long-context understanding, BestBrokers says the increasing sophistication of AI systems is driving a sharp rise in electricity demand.
According to the analysis, ChatGPT’s electricity use for answering user prompts has reached an estimated 17.23 terawatt-hours (TWh) per year, a level comparable to the annual power consumption of entire countries such as Puerto Rico, Slovenia or Iceland.
BestBrokers estimates that each ChatGPT query consumes around 18.9 watt-hours of electricity, more than 50 times the energy used by a standard Google search. With an estimated 810 million active weekly users submitting an average of 22 questions per week, the platform is calculated to process approximately 2.5 billion requests per day.
On this basis, ChatGPT is estimated to consume around 17.228 billion kilowatt-hours annually solely for responding to user prompts. Using the average U.S. commercial electricity rate of USD 0.141 per kWh, BestBrokers calculates the annual power cost at approximately USD 2.42 billion.
To illustrate the scale, the analysis suggests that ChatGPT’s yearly electricity consumption could power New York City for around 113 days, or supply electricity to the United States for about 34 hours. It could also power the United Kingdom for nearly 20 days.
The report further compares the estimated 17.23 TWh annual consumption with national electricity use worldwide. According to BestBrokers, this amount of energy could supply electricity to China for 15 hours, India for just over three days, Germany for more than 12 days, and France for nearly two weeks.
In household terms, BestBrokers estimates that the electricity ChatGPT uses annually could power all households in the United States for more than four and a half days, based on average daily household consumption. Alternatively, it would be sufficient to fully charge approximately 238 million electric vehicles, assuming an average battery capacity of 72.4 kWh.
Commenting on the findings, Alan Goldberg, data analyst and author at BestBrokers, points to a widening gap between efficiency gains and overall energy demand.
“Training runs for state-of-the-art systems now require tens of gigawatt-hours, and inference isn’t cheap either,” Goldberg says, adding that each query from the most advanced models can use significantly more power than earlier architectures.
He also highlights a lack of transparency around the true electricity footprint of large-scale AI systems, arguing that without clearer reporting and stronger optimisation standards, the infrastructure and environmental implications of rapid AI growth risk being underestimated.
BestBrokers notes that its full report includes detailed calculations, assumptions and source material underlying the estimates, as well as a broader examination of the power demands associated with large-scale artificial intelligence systems.
No related posts.
Editor-in-Chief • ScandAsia Publishing Co., Ltd. • Bangkok, Thailand
