Google Maps is relatively consistent level of quality across a country (some countries are much better mapped than others in GM), whereas in OSM one part of a city might be well mapped, and another part poorly mapped.
OSM better than GM:
Sarajevo: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlemap... POIs within Dublin: http://tools.geofabrik.de/mc/?mt0=mapnik&mt1=googlemap...
Regarding Dublin - your link has missing POI info, don't know why. Here's a screenshot directly from google maps page: http://cl.ly/1H2r0X2I1U2O3F0j0X2U
However, even without their contribution, OSM is already at a level of maturity to compete with the bulk of GM implementations. It seems that most people just need basic location services.
There are 2 problems here: (a) OSM's doesn't have as much house numbering data (that'll improve with time) and (b) addressing → lat/long is pretty freaking hard. If you limit yourself to one country and one address system, it's much easier. Otherwise things get messy fast.
At my last job I started to negotiate with Google about their (old) new maps pricing. They kept pointing to features that made no difference for our application. We needed Ok maps and decent forward and reverse geocoding. We were in their free tier, but needeed to find a solution for when we outgrew it, because their pricing was way way too high for our business model.
I left that company, bust if nothing else, this new pricing will further delay the day that they consider replacing Google maps. For new projects though, I'd probably skip google maps altogether, or at least use them in such a way that the client-side code would be totally provider independent.