This Engineer Built a Tool to Measure the Energy Cost of Your AI Chats

Ever stopped to wonder how much energy your AI prompt actually burns?

Probably not. Most of us don’t think twice about typing “Write me a blog post” or “Summarize this article” into a chatbot. But Hugging Face engineer Julien Delavande did — and he built a tool that does exactly that: calculates the energy footprint of your AI conversations.

Here’s the truth no one’s talking about — every prompt you send to an AI model eats up electricity. And with models running on heavy-duty GPUs and custom chips that guzzle power, those “small” interactions add up… fast. We’re not just talking about data center costs — we’re talking about a growing carbon footprint, query by query.

That’s where Delavande’s tool comes in. Designed to work with Chat UI (the open-source front-end for models like Meta’s Llama 3.3 70B and Google’s Gemma 3), the tool shows you — in real time — how much energy your messages are consuming. It even translates that into something tangible, like how long you could run a microwave or toaster instead.

For example, Asking Llama 3.3 to write a standard email uses ~0.18 Watt-hours. That’s about 0.12 seconds of microwave time. Multiply that by millions of prompts daily? You get the picture.

No, it’s not 100% precise. And no, Delavande’s not claiming to be the climate savior of AI. But this tool shifts the convo from “cool AI” to “conscious AI” — one data point at a time.

As Julien and the team put it: “One day, AI energy stats might be as visible as nutrition labels on food.”

We’re here for it. Transparency breeds better decisions — and AI that respects the planet might just start with knowing what your prompt costs.