Most people think artificial intelligence runs on electricity, but AI really runs on water, generally drinking water.
Every time you submit an AI prompt, upload a photograph, stream a video, conduct an internet search, or save information to the cloud, computers somewhere must process, transmit, and store that information. Those computers generate enormous amounts of heat and use enormous quantities of water, mostly drinking water, to keep them from overheating.
Artificial intelligence is real, not virtual. The cloud cannot exist without server farms. Data centers occupy real land, consume real electricity, and use real water. Artificial intelligence may appear to exist in a virtual world, but it depends upon very real physical facilities that require extensive cooling systems. Every discussion about the future of artificial intelligence must include a discussion about the resources required to support it.
Existing data centers generally rely upon evaporative cooling, where the water that transfers heat from the computers is sprayed into cooling towers and evaporates into the atmosphere to be lost forever until it rains again, usually somewhere far away. Air cooling, closed-loop liquid cooling systems, heat recovery technologies, and other methods can significantly reduce or eliminate big data center dependence on public drinking water for cooling. At present, however, those alternatives often cost more than using public drinking water.
Communities throughout the United States are being asked to conserve water, reduce consumption, and prepare for drought conditions. At the same time, many communities are providing drinking water to industrial-scale data centers whose primary purpose is generating private profit.
The question is whether private corporations should be permitted to consume public drinking water in pursuit of profits.
For extensive background information about this problem, see “The Insatiable Thirst of Big Data Centers”
AI Runs on water
June 7, 2026 | AI
Most people think artificial intelligence runs on electricity, but AI really runs on water, generally drinking water.
Every time you submit an AI prompt, upload a photograph, stream a video, conduct an internet search, or save information to the cloud, computers somewhere must process, transmit, and store that information. Those computers generate enormous amounts of heat and use enormous quantities of water, mostly drinking water, to keep them from overheating.
Artificial intelligence is real, not virtual. The cloud cannot exist without server farms. Data centers occupy real land, consume real electricity, and use real water. Artificial intelligence may appear to exist in a virtual world, but it depends upon very real physical facilities that require extensive cooling systems. Every discussion about the future of artificial intelligence must include a discussion about the resources required to support it.
Existing data centers generally rely upon evaporative cooling, where the water that transfers heat from the computers is sprayed into cooling towers and evaporates into the atmosphere to be lost forever until it rains again, usually somewhere far away. Air cooling, closed-loop liquid cooling systems, heat recovery technologies, and other methods can significantly reduce or eliminate big data center dependence on public drinking water for cooling. At present, however, those alternatives often cost more than using public drinking water.
Communities throughout the United States are being asked to conserve water, reduce consumption, and prepare for drought conditions. At the same time, many communities are providing drinking water to industrial-scale data centers whose primary purpose is generating private profit.
The question is whether private corporations should be permitted to consume public drinking water in pursuit of profits.
For extensive background information about this problem, see “The Insatiable Thirst of Big Data Centers”