How much water does AI consume? The public deserves to know – OECD.AI

Published date:

Resource Type:
Summary:

AI data centers consume massive amounts of fresh water through onsite cooling and offsite electricity generation, with training a single large language model like GPT-3 requiring millions of liters. This analysis details how big tech companies like Google and Microsoft have seen water consumption surge 20–34% year over year, creating growing tensions with local communities in drought-prone regions. The article argues that AI model developers must disclose water footprint data alongside carbon emissions and provides technical formulas for calculating water use — a useful reference for advocates pushing for transparency and accountability around data center water impacts.

Author:
Shaolei Ren
Organization:
OECD.AI

Comments

Leave a Reply