The Water Footprint Behind Artificial Intelligence
The dry season serves as a constant reminder of how precious water resources are. Amidst these conditions, farmers require water for irrigation, households rely on it for drinking, cooking, and sanitation, while the industrial sector and natural ecosystems, such as rivers and forests, depend on it to maintain the sustainability of life.
However, beyond these conventional needs, there is an entity that utilises water which is often unnoticed by the wider public: artificial intelligence (AI) technology. Generally, the public’s interaction with applications such as ChatGPT is perceived as a purely digital, fast-paced, and intangible activity, without any visible machinery, heat, or water. In reality, behind every query or question sent by a user, there operates a physical infrastructure consisting of massive-capacity computers within data centres.
Computers equipped with specialised chip processors consume large amounts of electricity during calculations, which naturally generates heat release within the devices. To prevent functional disruptions or damage to computers caused by overheating, data centres rely on cooling systems, whether they are air-based, water-based, specialised liquid-based, or advanced direct-to-chip cooling technologies that flow coolant directly to the hottest parts of the machine. This proves that AI is not merely a matter of algorithms and software, but is supported by physical infrastructure that is highly dependent on electricity and water.
In the operation of water-based cooling systems, issues regarding local hydrological dynamics arise. Although some water is recirculated repeatedly, some of it evaporates into the air during the heat dissipation process. While this evaporated water remains within the Earth’s hydrological cycle and eventually returns as rain, it does not necessarily return to its original location or the initial source of extraction. Consequently, the extraction of water from rivers or groundwater can deplete local water availability in the regions where data centres operate.
In addition to the issue of quantity, water quality issues also emerge. Water that is continuously circulated within cooling systems will experience a buildup of mineral content, meaning that some water (blowdown) must be discharged and replaced with fresh water. This waste water requires special management or treatment processes before being discharged, as it has a higher concentration of minerals and may contain chemicals used to support the performance of the cooling system.