A new analysis of artificial intelligence’s environmental footprint suggests that training and running models such as ChatGPT may already consume more water each year than humans drink from bottled water worldwide.
After combining estimated data center cooling needs with the largely opaque water-use data big tech companies choose to disclose, de Vries-Gao concludes that AI operations likely consume between 312.5 billion and 764.6 billion liters of water per year—bracketing and potentially surpassing the 446 billion liters of bottled water people drink globally.
Next Page
