During a recent conversation, a company representative stated that revealing specific total energy usage metrics is not feasible for various reasons. He mentioned that energy consumption is a fluctuating parameter, emphasizing the importance for users to contemplate their energy usage on a per-interaction basis. OpenAI publicly reports that ChatGPT handles approximately 2.5 billion queries daily. Using this figure and the company’s self-reported average energy consumption of 0.34 watt-hours per query, it can be estimated that annual energy usage totals over 300 gigawatt-hours. This figure is equivalent to the energy needed to power nearly 30,000 U.S. homes for a year.
Artificial intelligence (AI) extends beyond chatbots and permeates numerous aspects of daily life, often without users’ conscious awareness. AI-generated summaries can appear in web searches regardless of user intent, and email and messaging applications now feature tools for drafting and summarizing text. Google’s energy estimates, specifically for their Gemini applications, exclude other AI applications that the company employs. This complexity makes it increasingly challenging for users to accurately assess their personal energy demands related to AI usage.
While the speaker noted that users should not feel guilty for utilizing beneficial tools, he suggested that the broader discussion may not centered around individual responsibility concerning energy consumption. Instead, it raises questions about technology’s overall impact on energy use in society.
Source: https://www.technologyreview.com/2025/08/28/1122685/ai-energy-use-gemini/