It takes only a few seconds to create a picture. Just by typing a command, the appearance, clothes, or the atmosphere of the entire picture can change. But we do not know how much the huge infrastructure that works behind these few seconds of work costs. Not only electricity, but also a huge amount of water is required to use AI. With the increasing use of AI around the world, the cost of that water is now being considered. AI chatbots need huge data centers to write pictures or create other content.
The powerful processors there become very hot during work. During additional work, the temperature inside the processor can reach up to 176 degrees Fahrenheit. Many data centers depend on water to control this heat. Just as human sweat cools the body through evaporation, data centers are also cooled through evaporation of water. There are various calculations about the cost of water in using AI.
In the case of the GPT 3 model, the AI is given an estimate of using about 500 ml of water in a series of 20 to 50 medium-length questions. Gemini chatbot uses about five drops of water in a medium-length text prompt, so studies on AI's water usage vary widely because of different methods of counting. However, as AI's use increases, so does its demand for water. Creating images with AI is not an invisible magic. It requires powerful computer graphics processors and huge amounts of electricity.
Additional energy and water are also required to keep servers cool. However, the overall environmental impact of AI cannot be judged by the water consumption of creating an image. The main pressure is created when a large number of people create countless images, videos, and other content every day. The environmental impact of AI infrastructure is even wider. Minerals and natural resources are needed to manufacture computer chips and other devices.
Every step, from mining, production, transportation, and disposal after use, has an environmental impact. Data center power generation also requires water. Therefore, they are emphasizing the use of renewable energy such as solar and wind power. In addition, various technology companies are using liquid cooling methods.
In this, excess heat from the processor is absorbed using special liquids. The new generation of AI servers can be cooled with liquids as hot as 45 degrees Celsius. This can reduce the use of water to cool the data center. If the weather is suitable, it is also possible to remove heat from the data center through outside air. In addition, immersion cooling methods, in which components are directly immersed in a special liquid, are also being tested in various areas.