Want an energy efficient datacenter? Build it underwater

How submerged does this have to be?

Could periscopes aim optical networks, say from 20m below?

I'm imagining biomass consuming "creatures", now, with gills to draw the effluent waste from areas around raw sewage outflows that horrified me when little, despite ,my dad rarely corrected about such things, assured me they were legal at the end's distance from shore that i could easily spot nonetheless.

My optical networks knowledge is nil, but can simply larger diameter less pure plastic optic fibers transmit sufficient data, compensating by cross section in any way? Assuming one might try to build them strong enough to act also as physical ties, if other limits on movement like chains failed? Can such plastics be made?

Consider that there are many underwater volcanos. A data center would be negligible.

Recent AWS presentation showed off very less dense server configurations than any commercial vendor typically offers.

I guessed, i forget now if there was specific mention, that airflow simply costs a lot more than I first realized, after all i am only dumping host fan exhausts into someone else's co'lo design, which then has to deal with that. I bet it's substantially different, owning the whole building.

So, what if just submerging containers or pods of servers in cold sea water, combined with unforced airflow designs, was a passive cooling possibility?

After all, in theory, there's no surrounding facility to cool, and space is unlikely to be a major direct cost factor.

Despite I wonder, given land reliably increases in value, whether, once land acquisition is factored out, the kind of permits and early studies might bring costs back again.(excluding design NRE)

Marine insurance struck me as a important unknown.

I grew up by the Channel coast in England, and i can't imagine anyone allowing heavy containers to be anchored or moored where they might break free into shipping lanes, without quite detailed study and assurances.

Imagine if you built a data center was on a river that was fed by mountain snow melt, and you created a diversion that took some of that water, ran it along heat exchangers in your racks and the out through a hydro-power unit and back into the river on the other side. You might be able to build the first negative PUE data center in the world.

But you wouldn't tell anyone about it lest you lose your competitive advantage!

i am quite sure if most visitors here sat in front a randomly selected home computer, the relevancy of a message in the bottle, here, would be how fast pages render thanks to advertising networks. I took a 18 month mid-late sabbatical rolling up time owed, until last March, almost free of computing entirely, and I still cannot undo the effect i experienced, i drifted into rage ranting to an attorney friend how I had a claim for Theft Of Electricity (I think section 3 of Theft Act 80 in the E&W bit of UK) , i was so incapable of absorbing the enormity of influence.

(Disclaimer I work in advertising, but printed kind, on principle. I am naturally absolutely pro advertising, but Prof Mark Ritson is worth your time infinitely more than mine, to hear views not greatly distant, how advertising agencies have ruined their own business, making online ads a deeply revolting sick joke, on them as much as readers. Readers not "users" or even "visitors". Audience if it's audio or video, certainly, and for "reach" discussion of publications identifiable by social influence creating thereby a eagerly receptive audience willing to passively consume. But Readers we are. Prof. Ritson's talks on the tube are superb, and are better at teaching what I wish every customer, let alone every small customer* or startup. I beg anyone selling anything anywhere study what that man says about the industry you may rely on for customers and livelihood.

I still have Sting singing "Clickbait in a bottle" in my head...

* (my foible to say customer not client, because soliciting clients is unlawful in public, sorry bad law joke intimating prostitution by who seek @clients@ when most "clients" really want to buy concretely deliverable things)