Regular Joes like us can use AWS, GCE, on premises, some non-reseller colocation provider, etc., and create failover duplicates, alternative deploy targets, or simply not ever have a complete outage due to the unlikelihood of all of these things failing at once.