The GPUs, contrary the CPUs, are already massively parallel devices. On the CPUs, running a program simultanously on multiple ones, requires to make it parallel, which can be a significant re-factoring (and sometimes it is even impossible). On the GPUs, this problem doesn't exist: they already have a lot of internal processing units working parallel.
However, the chips still have manufacturing constraints. This is improving quickly (see Moore's Law), but as we can see, there are always a newest game which requires yet another VGA card invention. :-)
A single GPU can handle multiple displays, it is not a problem for that since a long time. Similarly, as a CPU can run multiple programs at once.
I can't see the reason, why would it needed to have multiple GPUs for the different displays.
However, if there is the manufacturing limit, then it can be feasible. You have to compare, which is better for you:
- Having a single, probably costlier GPU, handling both of the displays
- Handling the multiple displays with different GPUs
(1) may be too costly, or if your requirements are strong, this chip may be even not existing.