It's not at all a recent trend, it goes back at least as far as the early 2000s. Also, the obvious next step from thinking about processing a few tuples together, is transposing your view of DB tables all the way, to consider _columns_ rather than _tuples_. This makes a lot of sense when you're processing analytic queries rather than transactions, which typically modify individual tuples. And _that_ view goes back at least another decade earlier if not more.
See my own brief historical survey on the matter in the intro my monograph regarding a computational model for column-oriented analytic engines:
https://arxiv.org/abs/1904.12217
(this is a bit of a shameless self-plug because other, much more distinguished people than myself have written about this subject.)