AI agents can use roughly 600 times more energy than a simple chat prompt, according to The Decoder’s report. The gap comes from the way agents perform many model calls, tool uses, checks, and retries to complete a task that may look like a single request to the user.
That matters as companies move from chatbots to autonomous workflows. A prompt that asks for an answer usually requires one short interaction. An agent that researches, opens pages, writes code, validates output, and revises its work can multiply compute use behind the scenes.
The figure should be treated as workload-dependent rather than a universal constant. Still, it highlights an important limit: agentic AI may improve convenience, but its cost and energy footprint depend on how many hidden steps are required to deliver the result.