Spatial databases provides services to store and manipulate geometries, generally positioned in a geodetic system. The importance of the spacial database behind your GIS will mostly depend on the usage, but generally speaking, you can hardly talk of GIS if you don't have a proper spatial database for data storage.
Due to the fact that computers can only manipulate linear, one dimension data, you can split spatial databases in two logical parts :
- Geometry manipulation and indexing, with geodesy support
- Storage technology
The algorithms and logic used for the geometry manipulation are really specific, and then mapped to "classical" one dimension data to make them directly compatible with computers for storage. The only feature that have one foot in each world are the spatial-aware indexes, that uses algorithms similar to R-Trees.
For the storage, any underlying technology can fit, and won't change much the way you manipulate the spatial data. It might be a SQL database (and assimilated technology) or some kind of noSQL storage or something else. The main thing that will change is the spatial indexing, any other feature can be implemented with no major drawback (well except the occasional additional work).
So here is my conclusion : if you learn the way how to manipulate spatial data efficiently, and, depending on your ability to learn new technologies, you will be able to adapt whatever the technology is actually used. Learning the general concepts behind spatial data, especially for relational manipulation, is the hard part, and uses mature concepts that are not likely to change.