I've talked about procedural generation in the past and emphasized it's potential importance in the future. I've also discussed possible futures of the industry and what might be needed in those new environments. I still say that the future will be more focused on smarter development rather than simply adding more resources. To that end I would like to discuss specifically the limitations we'll face with our current forms of media and storage capacities. With the new power coming to us where we can create anything we can imagine and have it run in real time the problem becomes storage space. While it's true hard drives and their capacities are bigger and cheaper than ever and SSD's slowly replacing the whole market with their amazing speeds and ever increasing capacities we still are limited in things like discs such as blu ray which while it has an impressive amount of storage is also becoming obsolete. While I would appreciate things like delivering games on USB drives the fact remains digital delivery on PC is still going to be the primary method. If games like max payne are any sign of things to come, ignoring the fact they didn't bother to compress anything, I would say it's imporant to understand fully the power of procedural generation.
The insanely smaller file size and unlimited power of creation means you could for example remake the mass effect universe in a way that literally every habitable planet can be visited with many cities and a population for each with randomly generated missions and an almost infinite amount of gameplay. You could probably also reduce the installation size to a fraction of what it is without any loss in quality at all. Even if the entire game isn't done in such a way at least a hybrid could add substantially more content.
There's also the possibility of conversion technologies that can take your finished materials and analyze them and create algorithms for generating the same content while still cutting down on size. In addition other conversion tech out there might take your finished product and convert it to voxels and again reduce the overall file size without loss of quality. Not all voxel tech looks like minecraft, in fact some projects like euclidean have some very detailed end results. In particular euclidean claims you can model like you normally would and their conversion system takes care of all the work so you're not adding any time to your development cycle and don't have to spend much time learning a whole new system or method. In particular I see a bright future for voxels once they start catching on properly.
The only thing that I would say is clear at this point is that the future of gaming is growing in complexity and unless we want every single game that comes out to eat huge chunks of our hard drives then we have to find a way to cut file sizes where possible without sacrificing content.