Is AI bad for the environment?

By Kevin Naderian · Updated 17 November 2026

Short answer: Per prompt, everyday AI text chat is small (about 0.3 Wh). At global scale, AI is driving a real, fast-growing rise in data-centre electricity. Both things are true, and how we use and build AI matters.

The individual view

Google measured a median Gemini text prompt at 0.24 Wh of energy, about 0.03 g of CO₂e and 0.26 mL of water. Independent estimates for ChatGPT are similar (about 0.3 Wh). For one person sending dozens of text prompts a day, that's comparable to a few minutes of a TV or a short phone charge.

The global view

The International Energy Agency estimates data centres used about 415 TWh of electricity in 2024, roughly 1.5% of global electricity, and projects this to roughly double by 2030, with AI the biggest driver. That means new power plants, grid upgrades and water use for cooling in specific regions.

What actually helps

  • Use efficient models for everyday tasks. You rarely need the biggest model to draft an email.
  • Be intentional with heavy generation (images, video, long reasoning runs).
  • Support clean energy and companies that publish real numbers.
  • Support restoration, understanding it complements, not replaces, cutting emissions. See do tree-planting apps work?

Where SproutGPT fits

SproutGPT uses efficient open-weight models and keeps answers tight by default, then earns real trees as you use it. We don't claim this makes AI “carbon negative”. We think it's a better default, and we show our working on the impact page.

Keep reading

Start growing your first tree 🌱

Free every day, no sign-up. Your first answer is a few seconds away.