Large application requires direct, non‑NAT connectivity between workloads in all regions and on‑premises, but on‑premises already use all RFC 1918 ranges. What should you do?

The correct solution is to use non-RFC 1918 ranges with a single global VPC. This approach allows for direct, non-NAT connectivity across all regions and to on-premises because non-RFC 1918 (publicly routable) IP addresses are globally unique and do not conflict with existing on-premises RFC 1918 ranges. A single global VPC simplifies network management and routing. Using multiple VPC networks with VPC Network Peering would still require unique, non-overlapping IP ranges within each peered network, which is not possible if on-premises already consume all RFC 1918 ranges and you need direct connectivity. Overlapping RFC 1918 ranges, even with isolated VPC networks, would lead to routing conflicts and prevent direct, non-NAT connectivity, as the on-premises network would have identical address spaces. Adding Cloud NAT would solve the overlapping IP issue but violates the requirement for non-NAT connectivity.