Skip to main content
Shaolei Ren

Shaolei Ren

Nov 4, 2026

19:15
So I just wanted to better understand what goes into AI's water footprint and how does water usage further up the supply chain compare to water usage at the sort of end of the supply chain at the data centers?
19:27
The concept of AI as water footprint is very broad and includes the water used to mine the rare earth to manufacture the AI servers and also to produce the electricity.
19:39
And if you follow the carbon emission scoping definition, we have scope one, scope two, and scope three.
19:44
Scope one is the direct water consumption by the data center facilities.
19:48
And scope two is for generating electricity.
19:50
And scope three is everything else like recycling and manufacturing.
19:55
So typically, the scope three is the largest amount of water, but there's not much information available in the public.
22:00
We know that AI consumes a ton of water and a ton of energy, but do you believe that it can be optimized to basically reduce more than it's using, not only for its own use, but maybe even it could offset energy use or water use and other systems? Basically, do you think that AI can do more good than bad when it comes to the environmental impact? Yeah.
1:01
Why does a computer need water at all?
1:04
Well, because data center or these computers in general, they require a lot of energy for processing computing, like AI models or training or inference.
1:14
And this energy turns into heat, and we have to get rid of the heat to the outside.
1:19
And oftentimes, we are using closed-loop liquid-based cooling system to transfer the heat from the server to the outdoor heat exchanger.
1:28
And this process is a closed-loop system, and there is no water loss or water consumption.
1:35
But we have to further reject the heat to the ambient environment.
1:40
And during the winter time, this is a pretty easy task because you can just blow the cold air to take away the heat easily.
2:40
Is that true or false or somewhere in the middle?
6:37
Do you see, in terms of the science, any other ways to cool the data centers, that is, for the outside loop that you've talked about?
6:46
So there are several interesting approaches.
6:49
For example, waste heat recovery.
6:51
And this is more commonly used in Europe, where they pipe the heat to some district heating networks.
6:59
And in that case, there's no need to dissipate the heat, because the heat is just reused for district heating.
7:06
But that's much, much less common in the U.S. Alternatively, some proposals are using underground thermal energy storage to shift the heating load from the summer to the winter.
7:20
So that's also very efficient, but it just depends on the actual locality.
10:16
How does the availability of electric power affect power rates or power consumptions of other customers on the grid?

We value your privacy

We use cookies to understand how you use our platform and to improve your experience. Click “Accept All” to consent, or “Decline non-essential” to opt out of non-essential cookies. Read our Privacy Policy.