AI 代理使用的能量大约是简单聊天提示的 600 倍

Climate scientist Zeke Hausfather demonstrates in a new analysis that AI agents consume far more energy than the simple chat query figures reported by Google and OpenAI suggest. Over eight weeks of logging his use of the programming agent Claude Code, roughly 1,100 inputs generated over 14,000 model calls and 3.2 billion tokens, consuming about 170 kilowatt-hours of electricity, around 150 watt-hours per input. Projected over a full year, this level of intensive use produces CO₂ emissions comparable to running an electric clothes dryer. Hausfather sees the shift to clean energy sources for data centers as the most critical lever to cut emissions. Google and OpenAI have been reassuring users with low-energy consumption figures per AI query. Climate scientist Zeke Hausfather's Claude Code data tells a different story. Google says a median Gemini text prompt uses just 0.24 watt-hours (Wh), less energy than nine seconds of television. OpenAI CEO Sam Altman estimated the average ChatGPT query at 0.34 Wh , roughly on par with a Google search from 2009 . Both data points were published last year. Even then, it was clear the numbers misrepresent actual AI usage. They don't account for reasoning models that generate many times more tokens from a single chat query, multimodal processing across various formats, multi-agent systems, code generation, scaling across billions of uses, and plenty of other factors. Google's "study" in particular massively downplayed real energy consumption . Ad Climate scientist Zeke Hausfather has now calculated just how big that gap is in an analysis on The Climate Brink . One major blind spot remains: nobody outside AI labs knows the actual energy cost per token of a state-of-the-art model. Keep in mind that Hausfather's numbers are reasonable estimates, but still just estimates. Ad DEC_D_Incontent-1 They're also based on current usage patterns. AI labs are already working to scale agent-based systems that run autonomously on tasks for days, weeks, or even months . That could drive an exponential jump in energy consumption. Hausfather tracked his own usage in detail over eight weeks. Claude Code stores complete local logs of every session, including exact token counts reported by the API for each model call. Ad His 1,138 typed prompts triggered over 14,000 model calls, an average of twelve per prompt. Each prompt processed an average of 2.9 million tokens. For comparison, a typical chat exchange without reasoning or web searches runs about a thousand tokens. In total, his Claude Code processed 3.2 billion tokens. Of those, 96 percent were cache reads because at each of the 14,000 steps, the agent re-reads its entire accumulated context. The text Hausfather actually sees on screen, the model's output, accounts for just 0.4 percent of all processed tokens. Ad DEC_D_Incontent-2 His best estimate for total consumption comes to about 170 kWh of data center electricity over eight weeks, with an uncertainty range of 70 to 330 kWh. Hausfather directly measured only the token counts from his Claude Code logs, and the conversion to electricity values relies on three independent methods with different assumptions. Ad