As artificial intelligence spreads across nearly every corner of the digital economy, this does not go without environmental concerns: How much energy does it actually take to power this AI revolution? A new report estimates the electricity needed for ChatGPT as well as what this translates to as real-world energy costs.
To understand the true price of the massive use of artificial intelligence, the firm BestBrokers has provided a reasoned estimate of the number of prompts the chatbot processes every day. The estimate is based on publicly available user figures, average usage patterns, and assumptions about how frequently individuals interact with ChatGPT during a typical session.
By combining these inputs, the firm approximated the total number of prompts generated globally on a daily basis and then projected that activity across a full year.
The calculations show that the electricity needed for ChatGPT to process every single query by users (up to 3.2 billion queries a day, based on our estimate) is roughly 60.7 GWh a day, a massive amount of electricity, typically associated with large-scale industrial production, data centres, or manufacturing plants.
Over a year, ChatGPT uses around 22TWh for answering user prompts alone – the equivalent to the energy demands of a large city or a small country.
The scale of ChatGPT’s electricity use rivals that of entire nations. The 22 TWh of energy consumed annually to process prompts alone is enough to power roughly 2.1 million U.S. homes for a year. It also exceeds the annual electricity consumption of several smaller countries, highlighting how AI infrastructure is beginning to operate on a national-scale energy footprint.
The electricity used by ChatGPT to process user prompts every year is enough to power these countries for:
- China – 19 hours
- United States of America – 1 day and 20 hours
- India – 4 days
- Russian Federation – 6 days and 19 hours
- Japan – 7 days and 23 hours
- Brazil – 10 days and 15 hours
- South Korea – 12 days and 22 hours
- Canada – 12 days and 23 hours
- Germany – 15 days and 22 hours
- France – 17 days and 7 hours
ChatGPT now operates at an unprecedented global scale, with around 900 million weekly active users and an estimated 1.2-1.5 billion monthly users as of March 2026. If each user submits roughly 25 prompts per week, the chatbot processes about 3.2 billion queries every day, or more than one trillion prompts annually. This extraordinary level of usage makes the infrastructure behind AI assistants one of the fastest-growing sources of digital energy demand.
Answering those prompts alone requires vast amounts of electricity. Based on estimates for GPT-5 inference, ChatGPT consumes roughly 60.7 GWh of energy every day, or about 22.15 TWh per year. At the average U.S. commercial electricity rate of $0.136 per kWh, this translates to roughly $3.02 billion in annual energy costs just to generate responses to user queries.
In particular, individual AI prompts are far more energy-intensive than traditional search queries. Research suggests a medium-length ChatGPT prompt consumes around 18.9 watt-hours on average, compared with roughly 0.3 watt-hours for a typical Google search. In other words, generating an AI response can require over fifty times more electricity than a conventional web query.
