The fact that you can allocate virtual servers and use them as you please was never disputed. The debate was about the scope of applicability of such servers to running a game, as well as the economic feasibility of doing that.
They have made destruction physics asynchronous and are running them on servers. Running basic physics, like collision detection, on servers is a standard procedure in online games. So that in itself is not extremely surprising. Running more complex destruction physics asynchronously certainly required some development effort and is surely not trivial. But to put that achievement properly in context, the following question have to be answered.
- How much processing power is actually used for these calculations per collapsing mash?
- Are collapsing buildings reused for all connected players in a shared online match, and if so, between how many players are they shared?
- What was the development effort of building and testing such a physics component?
- How expensive is it to run these servers?
Looking at what is happening in these gifs, I strongly suspect that the processing resources required for a single collapsing structure are way less than "20x". As others have said, that collapsing mash just does not look "20x" more complicated than similar things seen in other games. You could likely reach that number by aggregating everything that is happening at the same time in that shared online world. But if that world is shared by, lets say, 20 players, the figure loses its impact. "20x" for 40 players is less of a talking point than "20x" per player.
The more important point, though, is the question of development efforts and server costs. As I argued in earlier threads, it makes no sense to build such a thing, unless your game absolutely needs it, or you explicitly want to have a tech demo. It increases development costs and it creates server costs. Since destruction is something that has obviously already been done on much weaker hardware, a developer would, under normal circumstances, just create a physics components that scales well with the local hardware that is available. It's easier, cheaper, and creates virtually the same game. My prediction is that, for these reasons, this will remain an insular case.