I think it is correct that, unless there are device technology breakthroughs, we won't see any order of magnitude improvements in PHY capabilities...at least from theoretical research. Increasing the data rate right now seems to be about increasing the availability of physical spectrum. Hence, the interest in cognitive radio and millimeter-waves. We've largely explored what MIMO can offer, although millimeter-wave communication does offer a domain for MIMO PHY research since many antennas are possible in small packages and the MIMO capabilities of the channel are not well established.
That being said, there are always going to be specific applications or specific design constraints that will require custom PHY design strategies, so be on the lookout for new applications. For example, recently there has been some PHY research on beamforming for millimeter waves. At millimeter-waves the beamforming is often not done at baseband to save hardware resources, which requires single-tap beamforming. However, the vast spectrum and substantial sample rates of millimeter-wave links (likely to be used in 5G systems) produces severe frequency selectivity. Hence, there has been some work on single-tap beamforming in a frequency selective channel, which is a difficult optimization to run at low complexity.
As far as higher layers, I'm not as excited as many others seem to be. I do think there will be substantial progress (especially in exploiting more information to configure networks more carefully, i.e., big data), but I don't think much of it will come from the academics or research institutions. I think it will likely happen through industry (standards-based) efforts when markets are at stake. Engineering/scientists/program managers are always looking for an elegant technology that changes the game quickly, but this area has been under a microscope for a while. I suspect that performance jumps in practice will happen through the slow, grinding process that results from prototyping efforts.
If you are looking for money to support your research efforts, you don't have much choice. Look at what your funding agencies are interested in (e.g., NSF, DARPA, etc.). That should tell you where to go....unless your name carries enough weight where you can define the areas of innovation on your own.