This is one of those scaling vs. switching cost problems. If you have a small town, then cars are a better option than transit. In a large metropolis, transit is better. The problem is that as a city scales up, each individual step results in continually adding infrastructure for cars without ever making the active decision to switch, and redesign the entire city.

When a small town grows into a medium, it makes sense to add car infrastructure rather than rebuild everything. Again when you grow from medium to large it remains cheaper to add that new interchange rather than change the fundamental structure of the city. Growing from large city to metropolis often means incorporating other cities and their car based infrastructure, and making a coordinated move to anything else is extremely difficult/expensive.

Edit: I doubt we could properly identify the "next big metropolis" early enough to put the proper infrastructure in place. Even if we could, the politics surrounding who gets the infrastructure spending would cloud the process. Neither do I think we could simply set up a new metropolis some where and "build it correctly from the start".

The best bet is to find a means of bootstrapping the conversion of existing metropolises to walkable/dense/transit oriented methods. How do we mortgage the immense future benefits to apply pressure to make changes now? Is there an achievable half step toward the goal that provides enough benefit to drive the rest of the conversion? When I'm confronted with a switching cost vs scaling cost problem in operations, I would be putting in dual use infrastructure. Maintain compatibility while removing dependency.