By: Haeley Cole-Boksner
Everyday, over 200 million queries are sent into ChatGPT. In order for the language model to process it, approximately 621.4 MWh are used per day. Per 100 words generated by ChatGPT, over one bottle of fresh water is used according to a research collaboration by the Washington Post and the University of California Riverside. This is just for processing queries and generating responses, while training AI models similar to ChatGPT requires around 700,000 liters of water.
Despite the gravity of AI’s environmental impact, it was proven to not have been the culprit of the recent California wildfires. Many social media users across all platforms have begun blaming ChatGPT’s water usage for the speculated lack of water to put out the fires, but experts say otherwise. The Association of California Water Agencies states, “Water supply has not hindered firefighting efforts. Reservoirs in California are at or above average storage levels for this time of year, thanks in part to years of proactive water management.” However, this misinformation and the subsequent disprovement has raised awareness of this matter.
DeepSeek, a Chinese startup, launched a new AI model that caused Nvidia’s biggest market loss since its founding just seven days after its launch on January 20th, 2025. DeepSeek’s new AI model is estimated to have a 92% lower carbon footprint as well as approximately 90% less energy consumption than ChatGPT.
If this new AI model is able to compete with ChatGPT in the long-term, a major reduction in the size of data centers could be on the horizon. This would result in a great decrease in water usage because ChatGPT’s data centers require such vast amounts of water to cool down due to their size, like how old computers used to overheat when used for too long. Whether customers are better off drinking a bottle of water and using their own intelligence or making the switch to DeepSeek is up to users to decide, but the future of the climate is in consumer hands.
Sources: