A new study is putting hard numbers on a growing concern: the AI systems designed to act on our behalf are far hungrier for electricity than the chatbots we've grown used to.
According to a report from finance.biggo.com, a study by researchers at KAIST (the Korea Advanced Institute of Science and Technology) found that AI agents consume up to 136 times more power per query than conventional chatbots.
The distinction matters. A chatbot typically answers a single question and stops. An AI "agent," by contrast, is built to complete multi-step tasks on its own — planning, searching, calling tools, and checking its own work. Each of those steps can mean another trip through a power-hungry AI model, and the study suggests those trips add up dramatically.
The 136-times figure, per the KAIST research as reported by finance.biggo.com, represents the upper end of the gap measured on a per-query basis. In other words, the more autonomous and complex the task, the steeper the energy cost compared with a simple back-and-forth conversation.
The source item does not detail the study's full methodology, the specific models tested, or the average (as opposed to maximum) power difference, so those specifics remain unclear from the available reporting.
Why it matters: The tech industry is racing to deploy AI agents that can book travel, write code, and manage workflows without human hand-holding — and this research is an early warning that scaling up that autonomy could carry a far larger electricity and environmental price tag than today's chatbots suggest.