ChatGPT's actual energy footprint per query is roughly one-tenth what the tech industry believed for years, but the total power consumption across all users remains enormous. In June 2025, OpenAI CEO Sam Altman settled a two-year debate by disclosing that an average ChatGPT query consumes about 0.34 watt-hours, roughly equivalent to what an oven draws in a single second. The widely cited figure of 3 watt-hours per query, which suggested ChatGPT used ten times more electricity than a Google search, turned out to be based on 2023 calculations using outdated hardware and overly generous assumptions.

The correction reveals a paradox that most coverage has failed to address: the per-query number is genuinely tiny, yet the aggregate impact is genuinely enormous. Both realities are true simultaneously. With ChatGPT processing roughly 2.5 billion messages per day, or about 29,000 messages every second, the platform consumes approximately 850 megawatt-hours daily, equivalent to the annual electricity usage of roughly 29,000 U.S. homes.

How Much Does Query Type Actually Matter for Energy Use?

The 0.34 watt-hour figure represents only a baseline. The actual energy consumption varies dramatically depending on what the model is doing. A standard text query sits at the low end, but more complex tasks demand exponentially more power. Understanding these variations is critical for anyone trying to grasp AI's true environmental footprint.

• Standard text queries: 0.24 to 0.34 watt-hours, roughly equivalent to a Google search and about one second of an oven's energy draw
• Reasoning models: Approximately 4 watt-hours per query, roughly 12 times more than a standard text prompt
• GPT-5-class models: Average around 18 watt-hours per query, representing a 50-fold increase over baseline ChatGPT queries
• Document analysis tasks: Processing a 100,000-token document reaches approximately 40 watt-hours, the highest category measured

This 130-fold range between the lowest and highest energy-intensive tasks explains why blanket statements about AI's power consumption often miss the mark. A user running a simple question through ChatGPT consumes far less electricity than someone using an advanced reasoning model to solve a complex problem.

What Does ChatGPT's Growth Mean for Global Power Grids?

ChatGPT's user base has exploded in recent years, reaching 900 million weekly active users by February 2026 and crossing 1 billion monthly active users on mobile in May 2026, making it the fastest app in history to hit that milestone. This unprecedented growth has direct implications for electricity demand worldwide. The International Energy Agency (IEA) projects that data center power consumption will grow from 415 terawatt-hours in 2024 to 945 terawatt-hours by 2030, and approximately 1,200 terawatt-hours by 2035.

AI's share of that total power consumption is expected to shift dramatically. Currently representing roughly 14 percent of global data center electricity use, AI is projected to account for 35 to 50 percent of all data center power by 2035. This acceleration reflects not just ChatGPT's growth, but the broader proliferation of AI systems across enterprise, consumer, and research applications.

How Does AI's Energy Use Compare to Bitcoin Mining?

A common comparison in tech circles pits AI data centers against Bitcoin mining operations, but the analogy breaks down in important ways. AI-era data centers already consume approximately three times as much electricity as Bitcoin mining, which uses between 138 and 175 terawatt-hours annually. However, the critical difference lies in flexibility: Bitcoin miners can theoretically switch off operations when the electrical grid needs relief, whereas AI data centers operate continuously to serve billions of users around the clock.

This distinction matters for grid stability. As AI consumption grows, the inflexibility of always-on data centers poses a different challenge than the voluntary curtailment possible in mining operations. Power utilities and grid operators will need to plan for sustained, non-negotiable demand from AI infrastructure rather than the variable load that mining presents.

Steps to Understanding Your Personal AI Energy Impact

• Calculate your daily query volume: Track how many ChatGPT messages you send per day. At 0.34 watt-hours per standard query, 10 daily prompts consume roughly 3.4 watt-hours, equivalent to leaving an LED light on for about 30 minutes
• Consider task complexity: Recognize that using reasoning models or document analysis features increases your energy footprint by 10 to 100 times compared to basic text queries, similar to the difference between a light bulb and a space heater
• Contextualize against other activities: A month of daily ChatGPT use consumes roughly as much electricity as streaming a single high-definition movie, making individual user impact negligible compared to aggregate platform consumption

While individual ChatGPT usage has minimal environmental impact, the aggregate effect across billions of users creates a substantial and growing demand on global power infrastructure. The challenge for OpenAI, other AI companies, and energy providers is managing this growth responsibly as AI becomes increasingly central to how people work, learn, and solve problems.