Do data centers need to use fresh drinking water or can they use water with some salt in it (I'm guessing not because the salt corrodes?). They could at least use gray water?

Data centers don’t need any water. They can reject heat directly to the air. If the price of water were high enough, that’s what they’d do instead of evaporative coolers. This is a fake problem driven by clickbait headlines or, at worse, mispricing of water.

> Data centers don’t need any water.

Is this using "need" to only mean "there is no possible alternative in any design"?

Your statement -implies- datacenters don't use water for cooling or, if they do, it isn't enough to cause any problem.

How do you square this with the fact that lots of datacenters DO use water for cooling? Yes, if water was prohibitively expensive businesses would find alternatives - that's what prohibitively expensive means so I'm not adding anything more than a tautology here.

Similarly does your claim that at worst this is a mispricing of water mean that they DO use water and it drives the cost up? Isn't that the point here, that businesses driving up the cost of water is bad for the people that also need (your definition of "need" above) water?

Water is a slightly cheaper way to reject heat if you have unlimited almost free water. I don’t think you’d use water for cooling if you had to pay residential water rates instead of like agricultural rates for water. It’s not at all a showstopper to just dump the heat into the air without using an evaporative cooler. It just costs somewhat more. Compared to the capital cost of an NVidia GPU server, the cost is small.

Ok, granted, I'm guessing they are just working with different capital costs (to build air-based cooling; i.e. via a huge air cooling tower). But I'm sure there is a place where data centers could be using subprime water for cooling economically without artificially low water prices (the issue then would be capital costs for transporting and using subprime water).

that was also my understanding. Nearby Mexico City, in a semi-desertic area (Queretaro), there are a high amount of data centers, Google, Microsoft and Amazon have a region there.

I would think they only need cheap electricity, connectivity and in the case of Mexico, far away from seismic activity.

Evaporating a gallon of water eats about 2.5 KWhr. Buying a gallon of municipal water costs me about 0.6¢, or about 2% of the equivalent amount of electricity. AC is efficient, so say we'd only have to increase water prices by about 500% to make it not worth it to use evaporative cooling.

Your claim is that increasing the price of a basic utility by several hundred percent to prevent a company from using a huge amount of it is a fake problem? That sounds like complete bullshit to me, and I don't even live in a city that's collapsing into the ground due to aquifer overuse.

If you are literally drawing so much from the aquifers that the city is collapsing into the ground, I think the water probably is mispriced.

And there’s no reason you have to charge industrial users of water less than residential users if the real price (meaning the cost to replace that water) is higher than the residential price. 0.6¢/gallon is less than I pay and I live in a wet city. I pay about 2¢/gallon. So for industrial electricity rates of 5-7¢/kWh, A/C is approximately a wash vs using residential water for cooling.

>If you are literally drawing so much from the aquifers that the city is collapsing into the ground, I think the water probably is mispriced.

Cool story. But every human needs water, and no data center needs water. Why should every person have to pay more to live just to prevent Amazon from breaking the world for a tiny increase in profits?

“And there’s no reason you have to charge industrial users of water less than residential users if the real price (meaning the cost to replace that water) is higher than the residential price” …is what I said. You can, of course, charge residential users less for the first 100 or 1000 gallons of monthly use, as is common for utilities in the US.

Good faith engagement on this topic probably requires thinking for 5 seconds if the objection you raised isn’t easily addressed by common methods already.

Additionally, industrial/industrial-scale-agricultural uses of water are almost always more sensitive to water prices. An incredibly poor population would benefit from the tax revenues and jobs offered by companies moving in.

I've heard Mexico City has a lot of leaking water pipes so updating that infrastructure might be better. Also to do desalination, they need to be close to the ocean. There are SMRs close to being in production (NuScale Power is one) but solar power might be another cheaper option.

CDMX is massive and very old. I can only imagine updating the infrastructure is going to be cost prohibitive. Still, it needs to be done.

Asianometry made a video on exactly this: https://youtu.be/aNNt9O70ASc?si=U2VSh5W3r6rfdA34

Desalination plants produce a lot of brine, which you need to dispose of. You can't just put it back in the sea without killing marine life. It's not a simple problem to solve.

The economics of desalination can be complex, but the TL;DR is that in most cases it's only economical when there's no alternative for potable water and not industrial/agricultural use (though that doesn't stop some places from doing it anyways for security or too much money reasons, particularly Saudi Arabia).

Desalination also doesn't help inland areas like Mexico city, as there'd also be large costs to pump the water from the coasts - Mexico City is 2200m above sea level.

The only nuclear powered surface vessels left in the US fleet are the Aircraft Super-carriers, and their water is created for supporting itself - ~5-6K sailors and any that's needed for operations (cleaning, steam launchers on the Nimitz, etc).

Nuclear micro-reactors are still a ways away and are probably not a viable option for developing countries - Mexico currently has only one nuclear power plant. There is probably a future for nuclear powered desalination for coastal cities in arid areas, but it won't help everywhere.

Is it actually cheaper to power this with nuclear plants (especially micro-reactors that don't really exist in production) than just deploying a lot of wind/solar/batteries to do it?

We don't know because nobody is mass building any nuclear reactors right now (micro or traditional - it's one of the main reasons there's so many cost overruns with anything being built right now). There's also no mass scale example of renewable desalination, either. It's almost all bespoke, so it's mostly theoretical.

If climate change continues and water becomes something to fight over like oil was in the 20th century, then we may finally see traction on both fronts.

But there are pretty large-scale deployments of renewables and battery storage going on, so presumably some fraction of the power used in today’s desalinization plants is generated from renewables (at least in China.) At very least there should be no deep mysteries about the generation costs.

Sure, but desalination is very power hungry. In Oman they managed to get a desalination plant to be 1/3 directly covered by solar in the desert and they still need alternative power for when that 1/3 is unavailable. Battery costs are still high. If it was that easy, it would be done.

What’s technically more achievable: adapting desalinization so it can work efficiently on intermittent power, reducing battery costs, or getting SMRs up and widely deployed along the Mexican coast in the next decade? I don’t have any idea if the answer to this but I don’t think they’re all equally pie in the sky either.

In theory, SMRs - *if* they're built out at scale (meaning for many other uses). The problem is the first users are going to pay a premium and it likely won't happen until they're already scaled out (and may still have to compete with renewable subsidies).

Renewables should still be competitive in certain cases, though.