As we all know, today, AI is everywhere. Everyone knows it. We use AI tools for our day-to-day work, editing photos, writing articles, generating videos and creating designs. You name it, AI can do it.

If you remember the time from 1996 to 2023, image generation, photo editing, and video animation were very difficult and expensive jobs. You needed special skills, expensive software, and powerful computers. But today? Anyone can type a few words into an AI prompt and get an amazing image or animation in seconds.

It feels magical. But here’s the side of the story we don’t often hear: all this AI processing needs huge supercomputers, and those supercomputers drink a lot of water. Yes water. And not just a little. Let’s talk about how and why.

1. Why AI Even Needs Water

I know, it sounds strange at first. We understand AI needs electricity. But why water?
Here’s the deal: when you process huge amounts of data like training ChatGPT or running millions of AI image generations, the servers inside data centers get extremely hot. To stop them from overheating, these centers use massive cooling systems.

And one of the fastest, cheapest ways to cool? Evaporative cooling. This method sprays or cycles water through systems to absorb heat, and in the process, that water evaporates into the air and is gone.

For example:

  • A 100-megawatt data center can use up to 2 million liters of water per day. That’s enough water for over 6,000 households — gone in one day, just for cooling computers.

2. The Real Numbers — How Much Water Are We Talking About?

Big tech companies are not shy in AI investment, but their water use is climbing fast.

  • Google: In 2022, used around 5.5 billion gallons (≈ 20.8 billion liters) of water. That’s a 22% jump from the year before. (AP News)

  • Microsoft: Used about 1.7 billion gallons (≈ 6.4 billion liters) in 2022, which is 34% more than in 2021. (Water Technologies)

  • Meta (Facebook): Also saw big increases, especially with their AI expansion.

And here’s something even more shocking — training a single AI model like GPT-3 can consume up to 700,000 liters of clean water just for one training cycle (arxiv.org).

Even using AI can add up:

  • A typical 20–50 message ChatGPT session may consume about 0.5 liters of water indirectly through cooling.

  • Multiply that by millions of daily users and you start to see the hidden cost.

3. Is This Really “Wasting” Water?

Here’s my take: yes, AI is putting huge pressure on water resources. Especially when you look at where the data centers are built. More than 60% of new AI-heavy centers are in water-stressed regions, places already struggling with drought.

Think about it: we’re taking clean, drinkable water, sending it through cooling systems, letting it evaporate into the air, and not reusing most of it. That’s a resource we can’t get back quickly.

So is it a waste? If you define waste as “using something in a way that isn’t efficient or sustainable,” then yes, it’s waste.

4. If That Water Was Used for Power…

Let’s have a bit of fun here. What if, instead of cooling AI servers, that same water was used to generate electricity through hydropower?

Now, water-to-electricity isn’t a straight swap here (because cooling water and hydropower work differently), but just to get an idea:

  • 1 cubic meter (1,000 liters) of water can generate roughly 1 kWh of electricity in some small hydropower systems.

  • If a data center uses 2 million liters a day, that could, in theory, generate 2,000 kWh a day, enough to power around 70 average homes for a day.

Again, not every liter can be used like this in real life, but the point is: the water could serve another important purpose.

5. The Bigger Picture — Electricity + Water

AI is not just thirsty; it’s also hungry for electricity.
Global AI workloads could use 85–134 terawatt-hours of electricity by 2027 — about the yearly power use of an entire medium-sized country. And when you generate electricity (especially in fossil-fuel plants), you also use water in the cooling process there too.

So AI is hitting water in two ways:

  1. Directly (data center cooling)

  2. Indirectly (electricity production)

6. Can We Fix This? Yes — And Some Already Are

Here’s the good news: this water doesn’t have to be lost forever. We can reuse it, and many companies already do.

a) Reuse and Recycle the Cooling Water

  • Microsoft & Quincy, Washington — Microsoft worked with the City of Quincy to build a water reuse facility that treats and recycles cooling water for its data center. This project saves about 138 million gallons (over 520 million liters) of drinking water every year (EPA.gov).

  • Amazon Web Services (AWS) — By 2030, AWS plans to use recycled wastewater for cooling in more than 120 data centers. Right now, about 20 facilities already do this, saving over 530 million gallons of potable water annually (DataCenterDynamics).

  • AirTrunk, Malaysia — Partnered with a local utility to run its data centers using treated wastewater instead of fresh water. It’s the largest project of its kind in the country (DataCenterDynamics).

  • Google, Singapore — Uses closed-loop cooling systems that recycle the same water continuously, saving 50–70% of fresh water compared to traditional methods (LinkedIn).

  • OVHcloud, Sydney — Runs a system that uses just one cup of water to cool servers for 10 hours, drastically reducing water waste (The Australian).

b) Build in the Right Places

Don’t put huge AI data centers in drought-prone areas. Place them where renewable energy and recycled water are available.

c) Smarter AI Scheduling

Run heavy AI processes during cooler parts of the day to reduce cooling needs. Spread workloads across multiple centers to avoid overloading one location.

d) Transparent Reporting

Tech companies should publish their water footprint alongside their carbon footprint so the public can see the real impact.

7. Why This Matters to Us

Some might think, “Well, I’m not running a data center, so why worry?” But every AI image, every ChatGPT conversation, every video you make using AI is part of this bigger system. The more demand we create, the more these thirsty supercomputers have to run.

If we don’t handle this wisely, communities could end up competing with AI servers for the same fresh water supply. That’s not a future we want.


So, while AI is shaping our future in incredible ways, we can’t ignore the amount of water it quietly sips in the background. Knowing the numbers and understanding the impact is the first step. The good news is that many companies are now exploring smarter cooling systems and ways to reuse this water instead of letting it go to waste. If AI is going to be the brain of our digital world, we also need to make sure it has a sustainable heartbeat—one that works hand in hand with nature, not against it.