The more data they have, the better-targeted advertising they can provide, because that data is more or less volunteered by their users even if conpanies don’t know to ask for it.
For most of the big companies, I’d say their UX is stuck in local maxima and can’t really become orders of magnitude better due to risk aversion — you’ll never collect the data that there’s a bigger maximum somewhere else that way. A smaller company willing to take risks can find those bigger maxima much more easily.
Just look at how much worse Google recently made the GMail UI.
Meanwhile, Apple has no scouts, a bad geocoder from OSM (compared to Google's), but notably better maps (on average) from all the footwork OSM does to import maps regularly from each of the thousanda of counties across the US.
Outside the US, countries like Germany literally use OSM as their offical map for plotting land ownership, utilities and numerous other functions.
Where Apple is dying is the lack of an Apple Scouts program and having massive teams in India essentially importing data from OSM, one other sketchy map vendor that is a decade out of date and user reports. The former should be automated, as Apple Maps India literally does not comply with the standards defined by Apple stateside, and there is no value being added in this process of manually importing map data.
I guess in a way it's similar to having capital. If you don't have any money, it's hard to start a business. Similarly, if you have no data, maybe in the future it will be hard to start a business as well.
There's still a long way to go to reach that point, but I think we are seeing the start of it. For example, it's pretty hard to build an alternative to GMaps/Apple Maps. Even if you have a novel take on how mapping data could be used, you'd still first need the data to build the technology around it.
How do you fight that? Should it be fought? Or should we just accept it as a new requirement of building a business? In addition to starting with a good team (i.e. industry experts) and having money, you now need to immediately start building as wide of a dataset as you can.