Wit IPFS you could rent a few cheap storage servers at different providers around the world and have them all pin your content. That should be resilient against most classes of failures (unlike a cloud provider where bad code pushed to production by the cloud hoster can cause failures in all availability regions).
It's not exactly the first protocol enabling that, but IPFS also has the benefit of distributing bandwidth if your clients all access the files via IPFS (since they will serve them from their cache).