Researchers identify the 'hidden energy cost' of AI agents for the first time

A KAIST research team has analyzed the computational cost and energy consumption of AI agents, which are systems capable of autonomous reasoning and action. The study reveals that these agents can consume up to 136.5 times more energy per query than conventional generative AI. While some reports highlight this figure as a viral peak rather than an average, the findings emphasize that data center power consumption is becoming a critical challenge as the era of AI agents begins.