|
|
|
|
|
by jawns
7 days ago
|
|
I am curious about the energy/expense of getting water that has been "consumed" by data centers ready for other use (e.g. drinking water), versus the energy/expense of getting water that has been "consumed" by residential users ready for the same use. To my understanding, the only thing that changes when water is used for data centers is its temperature. That's a lot different than residential use, where it's used in toilets and needs to undergo significant wastewater treatment to be cleaned enough to be re-used. So how do we compare apples to oranges for these very different use cases? Update: It appears my assumption about the only thing changing in the data center case is temperature. For much of that water, a phase change occurs (evaporative cooling), so it is no longer accessible to be recycled. |
|
* In air-cooled datacentres an approach is "direct evaporative cooling". They might take 30°C outside air and spray a fine mist into it, cooling it to perhaps 25°C before it enters the servers. After passing through the servers the air might leave at around 38°C. The water is now dispersed as humidity in a large volume of exhaust air. Recovering that water would require condensing it back out of the air, which means removing huge amounts of latent heat, it would be cheaper to just use 'traditional' compressor based cooling in the first place.
Cooling towers (which are used in many 'ai' facilities) have essentially the same problem. Servers reject heat into a water loop, and the cooling tower then cools that water by evaporating a portion of it into the atmosphere. The water that leaves as vapour is the "consumed" portion. While some liquid water remains in the system and a small amount is discharged as concentrated blowdown that can be treated and reused relatively easily, the majority of the consumed water has been converted into atmospheric moisture as above.