I know what you mean @Chris, it's one of those things that you are sure you should have thought of.
I absolutely love Project Natick and I've been following it for a couple of years, Naval Group who built and deployed it have some interesting videos on YouTube.
It's really not practical and definitely not sustainable to just keep on building more and bigger data centres the way we have over the last few decades. So, anything that offers alternatives to that is a great idea in my opinion.
So, it's a blindingly obvious solution in some ways. You take something that needs massive amounts of cooling and just drop it into a massive body of cold water, job done.
Of course then they need to get connectivity and power out to them but as I understand it they are only ever planned to be a pretty short distance from the shore.
These can, in theory at least, be deployed very quickly anywhere on the planet as there is a need, and with more and more edge applications there is a need for more 'local' data centres in places that they don't really exist right now.
There are really no issues at all until there is an issue. There is no such thing as a quick and easy fix if something breaks, so everything inside needs to be super stable and dependable, there needs to be plenty of redundancy, and with great remote monitoring, management and maintenance tools. That shouldn't be too hard though, I mean if we can remotely fix things in unmanned space then surely we should be able to do it n a box a couple of miles off the shore connected by a physical cable.
That physical lack of access means that the applications are a bit limited though, if you are in charge of this thing you are going to want as few random system admins and engineers poking around in stuff as possible, just because if something breaks it can't be fixed. So for companies like Microsoft, Amazon, Google, et al. it makes sense, but it is only going to make sense for companies who will fill the entire data centre themselves. They are also not upgradeable without pulling them up, so they are always going to be reliable rather than bleeding edge.
You aren't going to be able to co-lo in one of these for obvious reasons. Maybe that's a future idea though, bigger undersea data centres that actually do have access, either from the surface or by submersible/diving, where engineers can gain access. That would be pretty amazing!
Yes I saw this story too and thought it was pretty "cool" (groan). Reminded of a visit years ago when I was shown a data centre for Air Canada in Winnipeg Manitoba. I was shocked when we went up the elevator to take a look and it was on the top floor of a six story building. We walked into the centre and they showed me how they opened the roof to let the minus -35C cold air into the building and it was saving them a fortune in cooling costs. It's funny the obvious solutions that never occur to us.
Microsoft is all about innovating in this space recently.
Their new toy Azure MDC is a "data centre in a box". It's an RF shielded box that they claim will deal with whatever weather you can throw at it and has satellite comms, including using SpaceX's Starlink.
The idea is that they can be deployed anywhere (on land) for humanitarian, disaster relief, military, etc. operations.
I'm curious what kind of effects this would have on climate if deployed in greater scale. Obviously the equipment is dissipating heat into the ocean, which would raise ocean temperatures. We're already losing tons of coral due to rising temperatures, and coral reefs are where the greatest marine animal diversity is found. On the other hand, we wouldn't be running air conditioning to cool the data centers, which decreases carbon output. Would there be enough of an offset there to counteract the possible issues in the ocean? Would the effect in the ocean even be significant enough to be a concern? I don't have any real knowledge in this area.