In order to match with GCP Host project/service projects mechanisms (network resource is managed by the host project and used by its service projects), which is usefull for companies having a large number of GCP projects to manage, in different business units, it could be very usefull to be able to not manage VPC inside projects only, but also at Organization layer: an Atlas admin could create an "organization level VPC", and peer it with a GCP host project, and several Atlas projects could 'consume' this VPC, by running clusters in it.
That would permit to avoid having to create plenty of VPC (one per Atlas projects) and establish many peerings, what is technically limited anyway.