With the widespread adoption of artificial intelligence technologies, the energy and water consumption of data centers has become a global discussion topic. Asst. Prof. İnci Karakaş, stating that cooling systems and electricity generation processes, in particular, exert significant pressure on water resources and emphasizing that changing the energy source will provide water savings, said, “Renewable energy sources (solar, wind) are much more advantageous in terms of water consumption because they use almost no water operationally.”
Emphasizing that individuals also have a responsibility in AI usage, Dr. Karakaş stated, “It is necessary to reduce unnecessary queries and repetitions; instead of asking the same question multiple times with minor variations, a single, clear, and comprehensive question should be posed.”
Asst. Prof. İnci Karakaş, Deputy Director of the Institute of Health Sciences at Üsküdar University Vocational School of Health Services (SHMYO), explained recommendations to reduce water consumption in artificial intelligence (AI) usage.
Cooling technology directly affects water consumption
Asst. Prof. İnci Karakaş stated that the first recommendation for reducing water consumption in AI usage is “Making Data Center Cooling More Efficient,” and said:
“It is important to prefer technologies that reduce water demand, such as air cooling instead of liquid medium cooling, closed-loop systems, and ‘immersion cooling’. In the immersion cooling method, servers are completely submerged in a special non-conductive liquid. This system removes heat much more efficiently than air cooling; it ensures temperatures remain balanced and allows hardware to operate at high performance for a long time.”
The source of energy also determines water consumption
Asst. Prof. İnci Karakaş stated that the second recommendation is “Choosing the Energy Source,” and shared the following information:
“Coal and natural gas power plants operate with steam turbines, and after the turbine rotates, the steam needs to condense again. Large amounts of water are used for this condensation process. Instead, using renewable energy sources (solar, wind, etc.) significantly reduces indirect water consumption as they do not require cooling water during operation.”
Data centers should be established in low water stress regions
Asst. Prof. İnci Karakaş noted that the third recommendation is “Moving Data Center Locations to Low Water Stress Regions,” and said, “Instead of establishing intensive AI infrastructure in regions with limited water resources, it is effective to choose more water-secure locations.”
Asst. Prof. İnci Karakaş stated that the fourth topic is “Operational and Software Efficiency,” and emphasized, “Reducing unnecessary model runs, focusing only on queries with clear objectives and high value, and minimizing unnecessary AI interactions contribute to lowering overall water consumption.”
Asst. Prof. İnci Karakaş expressed that the fifth recommendation is “Increasing Cooling Efficiency,” and said, “Evaporative cooling, which is cooling done by evaporating water, is one of the most water-intensive methods. In contrast, methods like direct liquid cooling or immersion cooling significantly reduce water consumption because they do not require evaporation. While evaporative cooling uses a high amount of water, liquid cooling uses less, and immersion cooling provides very low water usage. Therefore, changing cooling technology significantly reduces water consumption.”
Changing the energy source provides water savings
Asst. Prof. İnci Karakaş stated that the sixth recommendation is “Changing the Energy Source,” and remarked, “Electricity generation is generally carried out by power plants cooled with water. Renewable energy sources (solar, wind) are much more advantageous in terms of water consumption because they use almost no water operationally. Therefore, not only cooling but also the source of energy affects water consumption.”
Asst. Prof. İnci Karakaş stated that the seventh recommendation is “Efficient Infrastructure Planning and Regional Water Conditions,” and said, “Establishing data centers in regions experiencing water scarcity increases the pressure on existing water resources.”
Policies and standards should be determined
Asst. Prof. İnci Karakaş noted that the eighth recommendation is “Policies and Standards,” and said, “With the policies to be determined, data center water usage should be transparently reported, water use efficiency should become standard, and new infrastructures should be directed to water-insensitive regions.”
Individual users can also contribute
Emphasizing that individuals also have a responsibility in AI usage, Asst. Prof. İnci Karakaş listed what can be done at an individual level as follows:
“Reduce unnecessary queries and repetitions; instead of asking the same question multiple times with minor variations, pose a single, clear, and comprehensive question.
Analyze long texts in one go, rather than in parts.
Reduce requests for large file processing.
Avoid unnecessary high-resolution visual and long video production.
Avoid requesting repeated visual revisions.
Perform data cleanup in cloud storage; delete unused large files and reduce unnecessary backups.”
Asst. Prof. İnci Karakaş added that preferring internet and hosting services that use renewable energy, if possible, would also contribute to water and energy savings.
