Because salt is corrosive, and the real estate is expensive.

Why not build in the cold north? Snow, ice breaking stuff, more expensive construction and work.

There’s a common misconception that these data centers are so big to literally suck up all resources… that’s not it.

It’s just corpos cheaping out.

Why the desert? Because evaporative cooling is cheap as heck, and low power, and works best in dry air. And the land is cheap. And grid energy is cheap.

Why local power plants and generators? Because it’s cheaper than grid energy; it cuts out the middle man. And it increases reliability. Not because there’s literally not grid capacity.

Hence, you are onto a thermodynamically interesting idea. The waste heat could be a “preheater” for desalination.

But of course they are not going to do that: it would cost more money.

Nor would they hook up the waste heat to local communities. Why would they pay to do that and extend construction time?

Also, as a counterpoint, osmotic desalination (which requires no heat) tends to be cheaper anyway, but is still a very, very expensive water source.

So fun fact: Microsoft played around with the idea of Underwater Data Centres and experimented with one off the coast of Orkney.

Here’s their in-house article about it, apparently they were pleased with the results.

Salt water is corrosive.

You may not have heard, but the ocean level is rising, not only posing a flooding risk, but more importantly drastically increasing the severity and frequency of water-based natural disasters right at the coast. So, okay, still flooding; but wind too, and sometimes circular wind.

Your DC needs to be in Nevada mountains with the salt mines and it needs to provide heat for the homeless at the air-cooling ejection ports, as the prophet William Gibson foretold.

Or build in cold climates where they could use geothermal for power and naturally cold air to bring the temps down.

I hear land in Detroit is cheap. As is Northern Canada.

Data centers are generally built in dry, geologically stable places with few severe storms and cheap power. Coastal areas typically check none of those boxes.

But, but, then they would produce at least something useful. No, that won’t happen.

Waste heat recovery is a thing, and the economics usually work out in your favor if the feed material is really hot. If it’s only mildly warm, you’ll need a lot of machinery to concentrate the heat and raise the temperature to a useful level. At some point, the investment just gets absurd and the idea gets scrapped.

Using heat as heat makes the most sense, since there are fewer steps where you lose some of the heat. Theoretically, you could boil water with server heat, but the massive investment is probably the reason why that isn’t happening everywhere. Running reverse osmosis probably won’t work, because you need electricity for the pumps, and converting heat into electricity comes with significant losses.

Both Google and Microsoft tried with actual underwater data centers. It looked like it was feasible, but repairing something when it’s deep underwater was not feasible.

Near water, there will be environmental issues – similar problems to factories and nuclear power plants, debris in intakes and warm water affecting marine life.

Basically, the solution is to build more efficient or less compute-intensive software, but that’s not where we all seem to be heading right now.

Shipping large volumes of water long distances is not generally feasible. Using saltwater to cool has a ton of extra complications as humidity and proximity to saltwater generally drives up costs for things needed to avoid corrosion.

Building in arid areas avoids a lot of costs, building near rivers in low humidity areas is even better.

https://www.scientificamerican.com/article/china-powers-ai-boom-with-undersea-data-centers/

oh don’t worry in 10 years the ocean floor will be full of DCs

Not enough heat. I think in Europe they want to use it for building heating, still not enough, they need normal heating too.