AI Agents Consume 600 Times More Energy Than Simple Chat Prompts

Zeke Hausfather tracked the energy consumption of the Claude Code AI model over eight weeks, revealing that AI agents use approximately 600 times more energy than a standard chat prompt. During this period, Hausfather noted that 3.2 billion tokens were processed, resulting in a total electricity usage of 170 kWh. He criticized the low energy figures reported by Google and OpenAI, stating they distort the reality of agent-based AI. This finding highlights the substantial energy demands of deploying AI agents, which may impact decisions on infrastructure and sustainability in AI development. For practitioners, this underscores the importance of considering energy efficiency in AI model deployment strategies. The Decoder reported.