An IPv6 address can be divided into two main parts, a "subnet prefix" and an "interface identifier".
One significant difference between an Ethernet network and an LTE network is that in LTE the link layer is point-to-point and two nodes on the link being the PGW and the UE, where as in ethernet it is not.
Now coming to your problem, In ethernet network subnet prefix can be shared by the four PCs each of which will need to generate a unique interface ID on the link /subnet in order to form a 128 bit globally unique unicast address. In LTE, PDN GW allocates to each individual UE a globally unique subnet prefix. The interface identifier (64 bits) is created by the UE and UE forms a 128 bit globally unique unicast address by combining the global subnet prefix and an interface ID. Since the subnet prefix is allocated to only one UE any interface ID it uses will suffice to ensure the global uniqueness of the full 128 bit address.
Let me know if looking something different.