I think people misunderstand why Kubernetes exists. It is the reverse OpenStack. Kubernetes has the potential to be the one unified API of the cloud. A middleware for proprietary cloud APIs. A few resources, like load balancers, are already at a point where you barely have to care about the underlying cloud provider. With operators and aggregated API servers (especially if they'll be offered as a service) provisioning resources could follow one well-known standard. Calling it now, within the next 2 years cloud database providers like Compose will offer a way to CRUD resources via CRD/apimachinery-compatible services. Few more years and we'll have a generic yaml spec for these resources that work out of the box on multiple providers (probably with a bunch of annotations that are vendor specific).