What does rendering have to do with the networking? The two are independent.
Networking is the way the clients communicate with each other and remain in sync. This will be your biggest concern.
Networking on phones has a wide range of performance. One minute a phone can be on an EDGE system getting around 8kb/s with very high latency, and then it can switch over to a nearby cell tower at 50Mb/s 4Glte speed. Connections may need to be re-established in the background without interfering too much with game play.
This will likely be your main consideration when it comes to the world size. How much of a world can you keep in sync at slow speeds? How much data will your world be modifying? How will you handle players that are in transit, frequently hopping between cell towers of differing speed?
The next biggest consideration will be the simulation (not the rendering).
The simulation is going to need to operate on a wide range of hardware. Of course, you can specify a minimum spec machine. Do you want to allow 1GHz Snapdragon and Hummingbird processors? Or do you want to require something like a dual-core or quad-core Tegra processor?
Assuming you are careful with your development you could create a different executable for different processors, giving better experiences to those with faster devices. After all, communications protocol is independent of processing. Again this is up to you.
Rendering is really the last consideration. Things that are simulated can easily be culled from rendering on a slower systems; extra graphical goodness and gratuitous particles can be rendered on high-spec machines.
Figure out your world and its communication needs first. Then improve the design of the communications system until it works for your target architecture. That will tell you how many people your world can support.