So first there is raw data collection. That is a pretty capital intensive process and its never really 'done' because the world changes a lot, so you need a data collection 'process', sort of like building a really complex software product that you can periodically 'tag' as a release and push out. The capital costs from either on the ground surveying, or aerial imagery. Its not acceptable to put a circle on the map and label it "Tokyo", there are streets and houses and buildings and parks. So the capital intensive part is taking aerial photographs, or satellite imagery, and post processing it into your tile layers. And that just gets you to one level of gross detail, there is so much more than that.
Because of the intensive capital investment folks who get map data understand they have a valuable bit of information. Its one of those things where it would cost big bucks to acquire it so you can sell it to others who want it.
Then there are points of interest, who are all the businesses in downtown? What are their hours? Are the streets one way? paved? gravel? private or public? Are there travel restrictions? All of this data can be gathered by on the ground vehicles but that takes time. (or a lot of money if you do it in parallel) and all that data has to be fused together into something which associates places with data. So think about how you would represent that vector from lat/long pair A to lat/long pair B was one way on weekdays and not on weekends?
One of the things that is really clear is that Google benefits Hugely from having their own in house search capability. That provides a treasure trove of points of interest. But did I mention search is hard? So any good mapping solution ideally has an in house search engine associated with it. Look at MapQuest, one of the originals, it was a search engine getting traction as a mapping application. Look at Microsoft, look at all the data they populate in their maps from Bing.
So there are at least three really really hard problems to solve before you can even worry about the user interface and how to interact with maps, data collection, data fusion, and web search. I'd be surprised if anyone felt they could just 'jump in' from a standing start and build a maps application in less than about 5 years these days.
There are already niche apps for things like public transport, though you may need an app per city/region. Even those niche apps aren't always accurate -- the NYC transit apps (2 or 3 different once) I tried while there earlier this year didn't pick up on station closures due to maintenance.
Maps/transit routing/etc is an interesting problem, just one that seems difficult to do right as a startup (on the scale of ios or google maps).
1. Apple Maps are introduced and nobody likes them. Google Maps are nixed and some bullshit reason is given for not letting them back into the ecosystem for a while.
2. A bunch of people see an opportunity to create something that Apple will buy. They drop everything and get busy.
3. 3-6 months later Google Maps are allowed back into the ecosystem. 99% of the newly minted startups go bust.
4. A year later Apple actually makes Apple Maps usable in iOS 7. The rest of the startups go bust.