This New Tool Tells You How Much Energy Your Chatbot Prompts Use

Ever stopped to think about the electricity your AI chatbot messages are quietly guzzling behind the scenes? Hugging Face engineer Julien Delavande did — and he built a tool to help quantify it.

Every interaction with an AI model — whether you're asking it to draft an email or just saying "thanks" — consumes energy. These models run on power-hungry GPUs and specialized hardware, and as AI adoption soars, so does its electricity demand. Experts warn that AI’s energy appetite could soon rival that of entire nations.

Delavande’s tool, which plugs into Chat UI, an open-source interface for models like Meta’s Llama 3.3 70B and Google’s Gemma 3, provides real-time estimates of energy consumption. It measures the electricity used per message in Watt-hours or Joules, and offers clever comparisons — like how your AI query stacks up against powering a microwave or LED light.

For example, a prompt asking Llama 3.3 70B to write a typical email uses about 0.1841 Watt-hours — roughly the same as running a microwave for 0.12 seconds, or turning on a toaster for 0.02 seconds.

While these are rough estimates, not lab-grade metrics, the goal isn’t scientific precision. Instead, Delavande hopes the tool sparks awareness and pushes users — and developers — to consider the environmental cost of everyday AI use.

“Even small energy savings can scale up across millions of queries,” Delavande and his collaborators wrote. “Model choice or output length can lead to major environmental impact.”

With rising concerns over AI’s carbon footprint, tools like this may help shift the conversation toward sustainability — especially within the open-source AI community. Delavande likens it to a future where energy transparency becomes the norm:

“One day, energy usage could be as visible as nutrition labels on food,” he said.

For now, his tool is a small but symbolic step in that direction — a reminder that even digital conversations have real-world costs.