Remember how your computer gets sweltering hot right after a day of constant use?
While we don't use huge amounts of water to douse that kind of heat, big computer factories do. And it is worsening the global state of water waste.
Annual Literacy Surveys. That's about 3 in 4 students using ChatGPT!
Now, imagine how overheated your computer will be with about three people connected and using the same PC.
But how does this translate to generative AI or ChatGPT harming the environment? And what's it to us?
How Does ChatGPT Damage The Environment?
Data Centers are facilities where tech companies store computer systems, applications, and other associated components to ensure the smooth processing of servers as you browse the internet, store your photos in a cloud, or ask chatGPT a question.
Every time you ask ChatGPT to generate a picture of a dog or to answer your homework, you are already relying on servers processed at a data center.
Picture those sci-fi spy films where the heroes sneak into rows and rows of blue shelves to hack into the main computer:
And ChatGPT, or AI in general, needs these data centers to work at optimal levels throughout the day. And for these data centers to work, they need to be cooled. They can't be overheating like your dad's old PC! For those data centers to be cooled, they're gonna need a lot more water than you might think.
Water Consumption
Just like how your computer's PC overheats when you use it for longer, Data Centers do too. And it needs to cool down before the components' heat could cause malfunction and corruption of critical data! That's where water comes into play.
Data centers use water to extract heat from overheating PC components, consuming nearly fifty centiliters of water, roughly a single plastic bottle in every 20-50 queries on ChatGPT!
To understand how data center water cooling works, let's first picture the basic components of a data center:
Image Credit: Lois Garga/Canva
CRAC units stand for Computer Room Air Conditioner. These units are used at data centers to suck in the heat generated by the server racks. These server racks are crucial computers where data processing involved in AI or generative AI takes place. The CRAC units then transfer the heat to a space under the elevated floor of a server room, as most data centers have elevated floors to facilitate the flow of cooler air underneath.
From there, cool air flows out through grills right below the hot server racks, maintaining the right temperature for the computer systems to continue optimal function. This is the basic process of data center cooling, but it is not quite the water cooling process yet.
“Water-cooled Chiller” Image Credit: Lois Gargar
The process starts with the component called the Chiller.
The Chiller transports chilled water (I blue) to the CRAC units to collect the heat it has sucked from the server racks. The now warmer water (yellow) flows out of the CRAC units and then returns to a component inside the chiller called the Evaporator.
In the evaporator, a refrigerant, or a working fluid found in most air conditioning units, collects the water's thermal energy and transports it below and to the condenser.
Additionally, within the condenser, there is another stream of much cooler water called the Condenser Water (II blue).
Condenser Water then mixes with the thermal energy (orange) and sends it to the cooling tower (black tank).
From there, the thermal water is sprayed on a stream of air (from the grills underneath the server racks) which removes the heat and cools the water down, thus becoming Condenser Water and returning to the condenser to collect more thermal energy and sending it back to the cooling tower again. The excess heat is then fanned out of the cooling tower, which is in effect the same as the heat we feel while passing by computers or in this instance, server racks.
Through this process, you can determine that artificial intelligence needs and consumes copious amounts of water to process data from 7 billion people all over the globe. Google, the ever-infamous search engine, consumed about 5.6 billion gallons of water in 2022 alone. The next year, that number only continued to heighten, reaching 6 billion gallons of water throughout Google's data centers worldwide.
But that's only for one artificial intelligence company. Globally, there are about 67,200 AI companies out there as of the year 2024. Even then, these companies normally have tons of data centers worldwide, the number already reaching up to approximately 10,978. But that's only up to December 2023, who knows how many more have popped up since then?
With the introduction of generative AIs like ChatGPT, the water demand has only increased further. ChatGPT even bests Google in the water waste category, zooming past with 25 times more contributions to water waste than the ever-famous search engine.
With all these data centers operating at full power due to the increasingly fast development rate of artificial intelligence, it wouldn't take a lot for us to realize just how much water these companies are going to need to cool their systems.
However, contrary to popular belief, water around the world is not going to run out. Water access would.
Want to Write for The Teen Magazine?
Share your ideas and get published on The Teen Magazine. Whether it’s entertainment, wellness, or academics, your voice matters here!
Apply NowHow AI Contributes to Worldwide Water Scarcity:
Image Credit: Ayano Tosin on Unsplash
Earth is not going to run out of water any time soon. It is present in the atmosphere, the seas are vast, and because of the water cycle, it is virtually impossible to actually run out of water. But the problem is that water is not accessible to most places in the world, especially to people who live in vast lands away from oceans, rivers, or any other bodies of water; Africa is an example.
Water scarcity also doesn't involve limited access to just any kind of water. Instead, it specifically pertains to limited access to clean, drinking water. About four million, or two-thirds of the world suffer from water scarcity at least once a month in a year. With how thirsty these AI data centers are especially with the rise of ChatGPT, it is only a matter of time before we start to question, every time we generate 20-50 queries on ChatGPT:
How about the child who could've drank that plastic bottle?
Take the Quiz: What Candle Scent Matches Your Overall Aesthetic?
Halloween is just around the corner. It's time to light that candle. Take this quiz and find out what candle scent suits you best.
What Can We Do About It?
The obvious answer is that we cannot let go of ChatGPT. Or AI for that matter. It has become so closely integrated with our lives that life without AI has become practically impossible. So what can we do about it?
Be An Advocate
Educate yourself on the processes that make ChatGPT harmful to the environment, and determine the possible solutions that we can do to make it better. For example, there have been proposals for a humidification system that can replace or do a better job than the traditional water cooling process. Look more into these solutions and use your social platforms or connections as a youth to enlighten other people on this crisis.
Read up on topics such as how firms can take up more sustainable water cooling processes, like finding ways to reuse water or even lessen energy costs. Influence your fellow youth on safer practices to mitigate the harmful effects of ChatGPT on the environment. Share solutions and have your voice be heard by fellow youth and big AI corporations.
Image Credit: Image Credits: further reading and enhancing your knowledge. Water is a crucial resource and it is needed by everyone, so don't hog it! Learn to broaden your mindset and be careful not to waste water in your own home. While this wouldn't directly solve or mitigate the water crisis in the context of artificial intelligence, it would definitely help the general problem. Image Credits:
#environment
#ai
#chatgpt
#data center
#harmful
#water waste
#water crisis solutions
#water pollution
#openaicom
#youth voices