The Energy Cost of Interacting With AI: An Insightful Analysis
Expressing gratitude to an AI chatbot after it answers a question might seem trivial, but the cumulative energy costs associated with these interactions can be significant. A recent tool introduced by Hugging Face offers estimations of the energy consumed during these engagements.
The ChatUI energy calculator provides real-time estimations of energy use while chatting with an AI model. It allows users to compare the energy expenditure to that of common household items like LED bulbs or mobile chargers. A user can either enter a specific question or choose from suggested queries, receiving both a response and an energy consumption estimate.
For instance, crafting a “professional email” took the AI just over 25 seconds, consuming approximately 0.5 watt-hours—equivalent to 2.67% of a mobile phone charge. Generating a 90-second script for transcribing software required about 1.4 watt-hours, which corresponds to 7.37% of a phone charge, about 22 minutes of running an LED bulb, or 0.6 seconds of microwave operation. Responding to a “thank you” request resulted in an energy use equivalent to just 0.2% of a phone charge.
It’s important to note that the estimates provided by ChatUI are approximations rather than precise calculations. The tool is compatible with various models, such as Meta’s Llama 3.3 70B and Google’s Gemma 3.
Comparing AI Energy Usage with Google Searches
Data from the International Energy Agency (IEA) reveals that a single request to ChatGPT consumes nearly ten times more electricity than a standard Google search—2.9 watt-hours compared to 0.2 watt-hours. If ChatGPT were to replace all 9 billion daily Google searches, it would result in nearly 10 terawatt-hours of additional electricity usage annually, which is on par with the energy consumption of approximately 1.5 million residents in the European Union.
The environmental impact of AI largely stems from the substantial power and water requirements needed to operate data centers. The IEA estimates that by 2026, the electricity consumption for AI could be tenfold compared to 2023. Furthermore, by 2027, the water consumption could surpass the annual water needs of Denmark as a whole.