Comment by geertj
4 hours ago
I don’t think people are confused between the two. What’s happening is that drive by objections have no real way to assess viability. That calculation you ask for has line items in it that greatly depend on engineering optimizations that, unless you work in the field, are hard to estimate accurately.
I don't think you really need those calculations to know that it is always going to be less viable than a traditional data centre. It's like the idea about putting data centres under water.
I'm sure you can do it and you can come up with estimates over how much it will cost... but it's always going to cost more than not putting your data centre underwater.
On those line items make the most optimistic assumption. E.g. use Elons Starship cost to orbit. Assume useful GPU lifetime of 6+ years. Assume max physical possible radiators etc.