When we think about the environmental impact of AI, we often think of energy and emissions. But there's another critical factor: water. Training and running AI models requires massive data centers, which rely on water for cooling.
How AI uses water
- Training large models (like GPT-4) can consume millions of liters of water—especially when cooling high-performance computing systems.
- Daily usage by millions of users also contributes to ongoing water demand, especially in regions that are already water-stressed.
Why it matters
Water is a finite and increasingly strained resource. Understanding the water demands of AI helps build more responsible infrastructure and inform policy, procurement, and use habits.
What can you do?
- Use AI efficiently: fewer queries and more optimized prompts.
- Support providers investing in sustainable cooling solutions.
- Push for transparency around AI water usage.