AI data centers require water to cool computer servers. photo by Jasmin Murdad
ChatGPT’s computer servers use water to cool following each calculation.
Supported in over 180 countries, ChatGPT is an AI platform used by students and professionals alike as a replacement for research. Deriving answers to questions, inspiration, and talking points from an AI platform is convenient because it consolidates research from a variety of sources; however, such quick and simple calculations are making negative impacts on the environment.
A GPT is a Generative Pre-trained Transformer, an Artificial Intelligence (AI) model that learns to deeply understand the thought process of humans to generate human-like text. The accuracy of ChatGPT’s human-like responses allows its users to pass off the responses of the platform as their own. ChatGPT is owned by OpenAI, an AI research and development company that was founded in November of 2015 as a nonprofit organization. Despite OpenAI’s investment from Microsoft, it remains an independent company governed by the OpenAI Nonprofit.
Since ChatGPT and other AI platforms generate responses in physical data centers, they create multiple environmental issues. A data center contains rows of computer servers stacked on each other with some cooling system capping the rows or running along the walls or ceilings. Data centers require water to cool the computer servers after each calculation, and for ChatGPT, calculations require over 500 milliliters of water for a 100 word response, according to Business Energy UK. 500 milliliters is equivalent to a standard bottle of water. The American Society of Civil Engineers (ASCE), breaks down how devastating this usage of water can be on communities. The ASCE illustrates how data centers like Google, similar to the ones used by OpenAI, require so much water that they end up “representing more than one-quarter of the town’s annual water consumption.” Water is a valuable resource and using such a large amount of water each time ChatGPT creates a response is unsustainable.
The majority of data center cooling devices require water including chilled water systems and evaporative cooling. However, more sustainable cooling methods have been developed and effectively cool computer servers at data centers. These methods include Calibrated Vectored Cooling and Computer Room Air Conditioning (which draws air to create airflow paths and draws air across refrigerant-filled cooling units). Simpler methods are also available, such as Free Cooling, which utilizes the atmosphere outside of the data center to provide the computer servers with new cool air instead of cooling the same air. Free cooling is also a very energy efficient option. Each of these alternate methods of cooling are discussed by DataCenters.com.
Although the use of ChatGPT allows its users to avoid completing work completely on their own, the practices of AI data centers result in the consumption of unsustainable amounts of water. If the data centers of big companies like OpenAI don’t prioritize using sustainable methods of cooling their computer servers, water will continue to be consumed in large and unsustainable quantities, potentially using water needed by the communities surrounding data centers.
