Yes, the dedicated servers might be less. But when one of them breaks, I have to wait for the provider to fix it. On EC2, I can replace it in 5 minutes.
With 75 good quality machines I bet you wouldn't see more than three hardware failure a year if that. All 3 of those would probably be drive failures. I imagine the redundancy of the software would handle that without a problem.
I agree that the continence of EC2 is very nice. We use EC2 a lot but just in a lot more "elastic" way. We have found it is cheaper to colo boxes if the demand is constant.