I think you're missing a few evolutionary pieces. Most data analytics systems that I'm aware of are not currently relational. Long ago, the data lived in memory, but memory was expensive, so everything was moved to disk. The relational model added the formalisms of normalization (to cut down on space, among other reasons), but the types of multi-dimensional queries used by the analytics apps required too many joins for this to work. So the data was de-normalized (eg. OLAP) to improve performance. As memory prices came down, people started putting the OLAP indexes and aggregates into memory to get a performance boost. Moving the data back to memory and returning to a normalized, relational model isn't so much "drastic new thing" as it is "logical next step". For me, the upsetting thing is that just as I'm getting good at the data warehousing thing, it seems we're going to be switching to being relational again.

Now, the PC can get back to what it really is: a Personal Computer. For the past several years, I've whined that PC development was being misdirected and bogged down by its mainstream attention. It's difficult to expect capitalism to produce OS's, applications, and technologies for a small crowd of engineers, techies, and geeks when there are masses of consumers out there willing to shell out tons of money for half-baked, inflated, useless garbage. "Your customers tell you what business you're in" and for the past several years, the PC industry has been in the business of "protecting" the user from their computers instead of "empowering" the user with their computers. This specialization of computing power is a very good thing and it's win/win. The masses of consumers get the features and applications they were looking for in the first place without dragging down a sophisticated, technical tool to be accessible to the lowest common denominator.