In my experience, a specific service in "The cloud" has an uptime similar to a single raspberry pi. Some services are more reliable, maybe upto three-nines.
If you design your systems as such - expecting them to fail, then using "the cloud" is fine. If a VM running in AWS goes down, fine, the ones in hertzner continue to work.
There's levels of availability you need -- what services can go down, in what geographic location, and for how long (do you need to go sub-second from all geographic locations? Because I'm not sure you can do that in any situations. On the other extreme are you happy with a TTL of 60 and failover to another IP, which barely counts as available in my book)
If you design your systems as such - expecting them to fail, then using "the cloud" is fine. If a VM running in AWS goes down, fine, the ones in hertzner continue to work.
There's levels of availability you need -- what services can go down, in what geographic location, and for how long (do you need to go sub-second from all geographic locations? Because I'm not sure you can do that in any situations. On the other extreme are you happy with a TTL of 60 and failover to another IP, which barely counts as available in my book)