Let's Talk About AI & Water Usage, Because It's More Than People Think

Note: This blog is NOT in defensive of AI. Please don't read it that way. I am simply explaining the viral post. Let’s talk about AI and water.

If you’ve been on social media lately, you’ve probably seen this claim:

“In 2025, AI used as much water as the entire bottled water industry.”

That is a massive statement. And when you see something that bold without a clear source or explanation of how it was calculated, I think you should be critical. So is it true?

There is no credible source for this. The honest answer is that it depends entirely on how the comparison is defined.

Yes, AI uses water.

Most of the water connected to AI is tied to cooling data centers.

AI models run on powerful servers inside large facilities packed with hardware. Those machines generate a lot of heat. To keep everything running safely, operators use cooling systems, and water is very good at absorbing that heat.

But this is where things get a bit more nuanced.

There are really two kinds of water use to think about.

The first is direct use inside data centers. Some facilities use evaporative cooling, which does consume freshwater as heat is released into the air. Others use closed loop systems that circulate the same water over and over with very little loss. Some even rely on reclaimed or non potable water instead of drinking water. So the actual impact can look very different depending on the technology and where the data center is located.

The second is indirect water use, which comes from how electricity is generated. If a data center pulls power from thermal power plants, those plants may use water for cooling too. When people calculate AI’s total water footprint, they often include this upstream water used in the energy system.

That is not wrong. In environmental research, it is common to include upstream impacts. If AI drives electricity demand, and that electricity relies on water intensive power sources, then AI is part of that chain.

The important part is consistency.

I am assuming many of the numbers regarding AI are inflated when they're doing their projections. There isn't much research at all into the subject.

If you include electricity related water for AI, you have to include it for whatever you are comparing it to. For example, the bottled water industry does not just use the water inside the bottle. It also uses water to produce plastic, manufacture packaging, run facilities, transport products, and generate electricity. A fair comparison needs to count those things on both sides.

If the same boundaries are used for both, the comparison can be reasonable. If they are not, the numbers can sound dramatic without being truly comparable.

Now think about scale.

The bottled water industry directly extracts and moves hundreds of billions of liters of water every year. That is a massive amount of freshwater being physically removed, bottled, and shipped around the world.

Could AI’s total water footprint reach similar levels? Possibly, but only if you define it broadly and include global electricity related water use. That does not automatically make the claim false. It just means the answer depends heavily on how the numbers were calculated, where the data centers are located, and what energy sources power them.

There is another piece that often gets lost.

AI runs on the same underlying infrastructure as many of the digital services we use every day. The same kinds of data centers, cooling systems, and power grids power streaming movies and TV shows, scrolling social media, storing photos in the cloud, online gaming, ecommerce transactions, and video calls.

If we are going to count indirect water use for AI, that same accounting logic should apply across this entire digital ecosystem.

None of this means AI has no environmental impact. It clearly does.

The issue is when indirect energy related water use is presented as if AI itself is directly extracting that full volume of freshwater, or when the assumptions behind the numbers are not clearly explained.

The better questions are more grounded and practical.

How much freshwater are specific data centers actually consuming?

Where are they located?

Are they using drinking water, reclaimed water, or closed loop systems?

What kind of power grid are they connected to?

Are they operating in water stressed regions?

Those are the kinds of questions that lead to better policy, smarter infrastructure decisions, and real accountability.

If sustainability is the goal, the focus should be on transparency, consistent measurement, cleaner energy sources, and continued improvements in cooling technology.

AI has an environmental footprint. So do the systems that support nearly everything we do online.

The real conversation is not whether AI uses water. It is how much, under what conditions, and how we reduce that impact over time.

That is the conversation worth having.