Water, Water Everywhere Nor a Drop to Cool ??

The Cloud is Thirsty: Why Your AI Model is Drinking Up Gallons of Freshwater

A few weeks ago, my friend -Dave -was proudly showing off his new AI-powered chatbot at a networking event. “It’s so advanced,” he said, “it can summarize books, draft emails, and even recommend the best pizza place based on your mood.” Impressed, I asked him how much it costs to run.

“Practically nothing,” Dave smirked.

I raised an eyebrow. “Sure, but have you thought about the water bill?”

Dave looked confused. And that’s the problem - most people don’t realize that behind the sleek, futuristic AI models and cloud services we love, there’s a hidden cost: water. Lots of it.

How Much Water Are We Talking?

Let’s break it down. Data centres - those massive warehouses full of humming servers - need constant cooling to prevent overheating. And guess what? The most effective way to cool them is with water.

Most data centres primarily use freshwater (sweet water) for cooling because it is more effective and causes less corrosion than seawater. While seawater cooling is possible, it requires expensive desalination or specialized infrastructure to prevent damage from salt and minerals, making it a less common choice.

Take Microsoft’s data centre in Quincy, Washington. In 2021, it used 1.7 billion gallons of water - enough to supply a small city. Google? Over 4.3 billion gallons in a year. And as AI adoption skyrockets, so does this demand. Training a single large language model can consume as much water as 300 Tesla cars being manufactured.

And it’s not just in the U.S. Malaysia’s AirTrunk recently announced a second data center campus in Johor, aiming to expand cloud computing capacity in Southeast Asia. While this will drive digital transformation, it also raises questions about sustainable resource management, as Johor has previously faced water supply challenges.

The Environmental Ripple Effect

For regions already facing water shortages, this is a big deal. Imagine living in Phoenix, Arizona - where water scarcity is real - and knowing that your late-night AI-powered Netflix recommendations are guzzling the same water that could have been used to keep your city hydrated. Not exactly a feel-good story.

And it’s not just about water consumption. Many data centers pull water from local rivers, lakes, and aquifers, affecting wildlife and ecosystems. Worse, some discharge heated water back into these sources, disrupting aquatic life.

So, What’s the Solution?

Companies are taking steps—Microsoft is aiming for water positivity by 2030, meaning they’ll replenish more water than they use. Google is investing in air cooling and recycled water. But is it enough?

As business leaders, we need to think beyond just energy efficiency and start considering water efficiency too. Sustainable AI isn’t just about carbon emissions—it’s about water footprints. If you’re running AI-powered applications, ask yourself:

• Where is my cloud provider getting its water?
• Are there ways to optimize my models to reduce computation and, by extension, water use?
• Should my organization invest in green data centres?

Let’s Make Smarter Tech, Not Just Smarter AI

Next time someone tells you AI runs on the cloud, remind them: the cloud is just someone else’s water supply. Let’s innovate responsibly.

What do you think? Are companies doing enough to tackle AI’s water consumption problem?