Starlink is useful, but people often confuse coverage with capacity density.
The rough hierarchy is: fixed fiber/cable > Wi-Fi > cellular > LEO satellite > traditional GEO satellite. Starlink is a huge improvement over old satellite internet, but it is still fundamentally a satellite system, not dense terrestrial infrastructure.
As a rough illustration, suppose one Starlink beam covers about 63 square miles and has 6 Gbps of usable downlink. At New York City density, that footprint contains about 1.85 million people. That works out to roughly 3 kbps per person. For comparison, network planners typically budget on the order of Mbps per person to comfortably support peak demand in dense urban environments.
You can improve that with more satellites, more beams, more spectrum, and better hardware. But even a 100x improvement only gets you to about 300 kbps/person. Additionally, this all adds cost... that isn't needed for the vast majority of time or space that the sattelites will be over. If the sattelites orbiting over New York also orbit over the Sahara Desert - every $ spent improving NY capacity is also improving the Sahara... but with no return.
The reason fiber, cable, Wi-Fi, and cellular work so well in cities is spatial reuse. Capacity can be reused block by block, building by building, apartment by apartment, tower by tower, and access point by access point. A beam from hundreds of kilometers overhead covers a much larger area.
So Starlink is excellent for rural areas, ships, aircraft, remote sites, disaster recovery, and backup connectivity. But it is not a replacement for terrestrial broadband in dense cities. It solves coverage much better than it solves urban capacity.