Summary

AI is very popular, it's incorporated into nearly everything, like AI overviews on google, AI visualization, picture generation, and much more. AI has been used by anyone and everyone on the internet; it's normal. However this district has impacts that many are ignorant to. AI consumes large amounts of water, amounts far too unsustainable to maintain with society's current water crisis. 1/4 of the world doesn't have access to safely managed drinking water, yet AI continues to increase its water intake, as of 2025 it was estimated to have consumed 765 billion liters.  The extremity of this situation cannot be stressed enough. This issue might not directly impact you now, however with the projected increase in demand of the data sector and as models improve, our supply will run dry, and you will be left thirsty.

 

AI Popularity

Evidently, AI is being used more frequently with 700-800 million active weekly users of ChatGPT in 2025. The data sector is growing and the demand for construction is rapidly increasing. From a technological standpoint this is great, it shows growth, expansion, and improvement for this sector.  However there is a negative impact that often goes unseen.

 

Overview

AI is consuming unsustainable amounts of water as the global system was estimated to have consumed 765 billion liters in 2025. This amount is greater than that of the total bottled water consumed in the same year. Putting into perspective the extremity of this issue. Big tech companies often locate data centers in drought stricken areas to source water for cheap, sucking our lands dry. Water is a scarce resource and should be treated as such, especially in a time of global water bankruptcy.

 

AI Advancement

AI will continue to develop rapidly, which has a direct impact on  how much water is consumed. The more advanced the model, the steeper the impact is on water consumption. These complex systems overheat much faster since there is so much more coding for machinery to register. For example GPT-4 uses around 500ml (17oz) for the generation of a singular 100 word email. A 10 - 50 query chain can use up to 2 liters (1/2 gallon). The cost of using AI is the consumption of limited natural resources, is this trade off truly worth it? Is AI more important than rivers running from bank to bank? 

 

Causes

AI is a very physical thing. It lives in data centers all across the world. These machines are running 24/7, so anyone no matter location or time can access them. All interactions with bots are directly accessing these machines. Every little bit of computing will generate heat. Heat is very bad for these machines as overtime it can damage and ruin their efficiency.  Data sectors have to cool these machines in order to prevent any issues. Many data centers used to use air conditioning, like the features of a Macbook Air when it overheats. However the demand for technology and AI is much greater now, so the setup of data centers had to change, so companies pivoted. To fit more tech into the same spaces they made everything very dense, tight, and close. This made it impossible for cool air to breach these tight areas so they had to use water, which can absorb 4000 times more energy per unit volume than air. 

 

Efficiency of Data Center Cooling Systems

A very common cooling system is the open loop cooling system. This means that the water used to cool tech makes direct contact with air. This allows the water to evaporate into the atmosphere, and increases contamination. To clean the contaminated water that gets dust, debris, and others in it, chemical treatments are used. Things such as biocides, corrosion, and pH adjustors are introduced. Blowdown water from cooling towers can spread the chemical residue from discarded water to the environment. This can contaminate more water sources or habitats. This system is often applied to data centers since it has a lower initial cost and simple design that makes maintenance and construction easier. 

 

Location of Data Centers 

Companies like cost efficient choices, so they often pick areas with cheap land and water. On top of this, technology prefers dry climates in order to avoid rusting and damage from humidity. These insatiable entities are placed in previously scarce areas; do you see the issue with this? Since all data centers have similar preferences, it causes high density in certain areas. For example, Virginia has the highest data density in the world with 600+ data centers locating in state. Some places have had to start turning data center construction down due to water strain. These hot and dry areas aren't very beneficial for the temperature of AI either. If some place is hot, it'll translate to the systems. This location choice makes machinery overheat quicker, resulting in the use of more water to keep them tech functioning and safe from need of repair. 

 

What Normal Civilians Can Do

The best thing is to stop using AI. That stops all computing, and since computing is what generates the heat, there wouldn't be any water necessary to cool. However AI is impossible to avoid and is everywhere, so here are some simple things to do :

• Limit personal AI use by downloading Ecosia (AI overview is a toggle - you're able to turn it off or on)
• Don't use AI when not necessary and when it is used make your prompts very good and specific to get the response you want right away
• Educate others, some people use AI without knowing its impacts. If you inform them it could urge them to use the tool more conservatively
• Fight back - don't let AI poach our resources and stay hydrated
• Usage speaks words - so if no one is using AI, the data sector will be forced to make a change to their systems

Different Rules that Could be Applied to AI Usage

Some people might know this information but not care. So some restrictions of AI could be beneficial too.

• 1-2 AI questions allowed per day. This will encourage people to really think about their prompts and limit usage
• Make all AI features a toggle that can be turned off

Alternative Systems to More Efficiently Use Water in Data Centers

• The closed loop cooling system is a system where the water used to cool is in sealed pipes. The water makes very minimal contact with air directly. This prevents contamination and the use of chemical treatments. It also prevents the water from evaporating into the atmosphere. Comparatively, Closed loop conserves up to 70% more water than of open loop cooling systems.  This system is a massive improvement when it comes to conservation of water. However it is not as commonly implemented because of the increased complexity of construction and higher initial cost. A lot of decisions are focused on the cost of systems. However the cost is a sacrifice that needs to be made in order to prevent further water scarcity.
• Underwater data centers. Currently there is a big issue with underwater data centers and security, but in the future when soundproofing technology improves these would be great to implement into our data sector. The ocean has vast amounts of water that can be used for the cooling of systems. Passive heat transfer can cause 40% to 60% reductions. This intentional location choice takes advantage of the massive amounts of ocean water there is. Conserving the increasingly limited freshwater. 

Conclusion

This issue is projected to grow fast, our supply will be drained before we even know it. Something has to change, and it has to happen fast. 1/4 of the world already lacks access to clean drinking water, and that number is already large enough. If AI is taking priority over those dying of thirst, what makes you think that it would be different when it comes to you? Don't let AI take priority over our resources. Change is possible, but people need to make an effort and encourage the change. It won't happen on its own.