The way the GCP does networking is very different from what most networking professionals are used to, and other cloud providers such as AWS. This makes GCP networking confusing, especially since some of the important terms, such as VPC and subnet, have somewhat different meanings than in other contexts:

There is a lot going on here, so let's go through and parse the implications of all of that:

Each project in a GCP organization contains at least one default VPC. It can contain more than one network, depending on the resources requested. All VPCs of GCP are global resources, which means they are not bound to any particular zone or region. VPC resources have both internal and external IPs. The resources within the network communicate with each other using internal IPs, whereas external requests are served through external IP addresses...