Perfection? No; probably never.
It's not difficult to point to cloud outages and failures. It's going to happen. The hardware routinely fails, and although there are some remarkable coping strategies (Amazon's S3 service claims 99.999999999% durability and 99.99% availability, due to intra-centre and cross-data-centre redundant storage and retrieval), it's a fool who assumes their data and servers don't need a backup and failover strategy.
If you need just one server, cloud might very well not be for you, unless you're smart about backup. It's more useful to corporates who can commoditise their services into customised server images, and just spin up more servers (maybe in other data centres/availability zones) when things go pear-shaped. The use of edge caches and putting the whole infrastructure behind edge routers that can redirect to still-working centres allows service to continue unabated, in the face of large-scale failures.
Just please don't enter into cloud initiatives wearing rose-tinted specs. Things will fail. You do need to analyse and correct for each possible single point of failure, to maintain backups (ideally including at least one off-provider route), and to manage your cloud server and storage estate actively - so don't go thinking you can sack most of your BOFHs. You still need them; they'll be just as busy managing your virtual estate - but you can reduce your machine room CapEx to a tiny amount, and concentrate on pay-as-you-need OpEx.