How Much Energy Does AI Use? What You Can Do About It

From writing emails and generating code to answering questions and helping businesses grow, AI tools like ChatGPT are becoming part of our everyday lives. And while the benefits are clear, the energy costs behind the scenes are often overlooked.

‍

Running large language models requires massive computing power, which in turn requires electricity. Every time a prompt is entered or a response is generated, it happens in a data center filled with powerful servers running 24/7. Those servers need electricity not only to compute but also to stay cool. Multiply that by millions of users and billions of queries, and the energy use adds up quickly.

‍

In fact, a single ChatGPT query uses around 0.34 watt-hours of electricity, roughly enough to power a low-energy LED bulb for a few minutes or run an oven for a single second, according to OpenAI CEO Sam Altman (Business Insider, 2025). That may sound small in isolation, but when scaled to hundreds of millions of prompts a day, it becomes substantial. Researchers estimate that if usage continues to grow at its current pace, the daily energy required for AI queries alone could match the annual consumption of over 35,000 U.S. homes.

‍

Even though the impact of a single prompt may seem minor, it adds up quickly. Some studies suggest that training a large language model like GPT-4 can consume as much electricity as several hundred U.S. homes use in an entire year. And that’s just the training phase. Day-to-day use continues to draw energy and resources indefinitely.

To put this into perspective:

• One AI prompt emits about 0.05 to 0.10 grams of CO₂, about 4 to 5 times more than a typical Google search.
  
  
• One prompt uses approximately 0.34 to 0.43 watt-hours, depending on the model version and length of the input.
  
  
• Over time, this could amount to significant emissions, 10,000 prompts a month could equate to the emissions of driving 4 to 5 kilometers in a petrol car.
  
  

What’s more, AI’s environmental cost isn’t just about electricity. It also includes water. Every AI query consumes a small amount of water for cooling purposes, around 1/15th of a teaspoon per ChatGPT interaction.

But this isn’t about blame. It’s about awareness and action.

AI can, and should, be part of the climate solution. It can help optimize renewable energy, track deforestation, and improve energy efficiency. But like any powerful technology, it has a footprint. And understanding that footprint is the first step to managing it responsibly.

‍

At ON A MISSION, we’re exploring ways to help users and companies take meaningful action around their digital climate impact. Through our “Plant for Prompts” initiative, we’re giving individuals and companies the chance to compensate for the emissions associated with their AI usage by funding community-led reforestation projects in areas that need it most.

‍

Because yes, AI is powerful. But so is planting a tree.

When we connect the two, tech and nature, we can create a future where innovation and regeneration go hand in hand. One that recognises our impact and acts on it.

So go ahead and use AI to work smarter, learn faster, and build better. And when you're ready to give something back to the planet, we'll be here, planting trees that grow alongside your progress.