There is provided a system that includes a local cloud radio access network (RAN), and a remote cloud RAN. The local cloud RAN processes latency-sensitive applications, and the remote cloud RAN processes latency-tolerant applications. User traffic is appropriately routed to the correct cloud RAN based on the application. User equipment (UE) has no knowledge of which network is being used for processing, i.e., this network processing split is done in a manner that is transparent to the UE, e.g., by dynamically selecting a different access point name for local vs. remote processing. The processing split of the RAN between the local cloud RAN and the remote cloud RAN is done in a dynamic manner depending on the number of devices requiring low latency support. This allows the local cloud RAN to be very compact and low-cost since it does not have to process the latency-tolerant traffic.