# AI, Energy, Water, and Climate

AI's compute footprint is a growing environmental and policy issue. This sub-article expands [Section 1.5 of Hardware & Software](../hardware-software.md) and [Economic & Social Impacts](economic-social-impact.md).

## Energy Demand
- Training frontier models consumes megawatt-scale power.
- Inference at scale adds continuous load; data-center demand is rising.

## Water and Cooling
- Liquid cooling and evaporative systems consume large amounts of water.
- Drought-prone regions face tension between compute growth and water supply.

## Efficiency and Carbon
- Efficiency research (quantization, distillation, sparse attention) cuts cost per token.
- Renewable energy and carbon accounting are emerging priorities.

## Policy and Transparency
- Some jurisdictions require reporting of energy/water use.
- Compute governance debates include environmental criteria.

## Why It Matters
AI's growth is colliding with climate goals; efficiency and transparency are essential.

## Sources
- International Energy Agency. *AI and energy reports.*
- Data-center water studies (e.g., Li et al., 2024).

*Back to [Hardware & Software](../hardware-software.md) · See [Frontier AI](frontier-ai.md).*