One Email With ChatGPT Equals to One Bottle of Water

According to research conducted by The Washington Post in collaboration with the University of California, Riverside, generating a single 100-word email using ChatGPT powered by GPT-4 consumes roughly 519 milliliters of water. This amount slightly exceeds the volume of a standard 16.9-ounce water bottle. The reason behind this resource usage lies in the energy-intensive process required to run and cool the massive data centers that power AI models like GPT-4. The extravagant consumption of water, particularly in such small interactions as writing an email, contributes to the broader environmental challenges we face, including exacerbating human-caused drought conditions in regions already struggling with water scarcity.

The reliance on water in the operation of generative AI servers stems from their dependence on electricity. These servers require a substantial amount of electrical power to function effectively, as they perform complex computations and process vast amounts of data. However, generating this electricity often involves considerable resource consumption, including water.

The cooling systems essential for maintaining optimal temperatures within data centers utilize water to prevent servers from overheating. As servers operate continuously, they generate significant heat, and without adequate cooling, their performance can degrade, potentially leading to failures or slowdowns. This necessity for cooling water means that, alongside the energy demands, the operational infrastructure of AI technologies like ChatGPT also significantly impacts water resources.

The cumulative effect of millions of emails sent daily can lead to a considerable environmental toll. As more people turn to AI-driven tools for communication, the energy demands associated with processing these emails increase. This rise in energy consumption not only contributes to greenhouse gas emissions but also places additional strain on energy resources, potentially exacerbating climate change.

In areas already facing water scarcity, this combination of high electricity consumption and water usage can create pressing environmental challenges. The dual reliance on these resources underscores the importance of considering the broader implications of deploying advanced AI systems, particularly in regions where water conservation is critical. As we advance technologically, finding sustainable solutions for energy and cooling needs will be essential to mitigate these impacts

Other findings include:

• If one in 10 working Americans (about 16 million people) writes a single 100-word email with ChatGPT weekly for a year, the AI will require 435,235,476 liters of water. That number is roughly equivalent to all of the water consumed in Rhode Island over a day and a half.
• Sending a 100-word email with GPT-4 takes 0.14 kilowatt-hours (kWh) of electricity, which The Washington Post points out is equivalent to leaving 14 LED light bulbs on for one hour.
• If one in 10 working Americans writes a single 100-word email with ChatGPT weekly for a year, the AI will draw 121,517 megawatt-hours (MWh) of electricity. That’s the same amount of electricity consumed by all Washington D.C. households for 20 days.
• Training GPT-3 took 700,000 liters of water. (source)