Even if priced against data centre-type hardware, it's still an insane mark-up.
It's a typical trick of an 'unbundled' service where you find out long after the buy-in point (and when there are serious expenses involved in moving) that it is more expensive than you thought after all. That's why I still maintain that regular, dedicated hosting is unbeatable both for high storage and high bandwidth solutions. If you just want to play in the pool you might get a better deal using virtual servers but as soon as things get a bit more serious you will probably want to move. The same goes for 'the cloud'.
There are good use cases for this stuff but there are plenty of online concepts for which they are not a good match and way too expensive for what you get.
And redundancy does not only increase the cost of the physical, it also increases maintenance, since the devices fail and have to be replaced. Also, we are not talking hard disks here. If a server, memory, processor or anything else fails, it will be replaced and your virtual machine (and all others that happen to run on the failed hardware) will be migrated to another chassis so it can continue running. If they have supernaturally high availability, it could even be running in sync on two different boxes with smooth handover in case one fails and you won't even notice there was a glitch.
But I guess that's too much to ask yet. I am not aware this kind of technology has been developed.