Sadly, spectrum is a scarce resource. The lower frequencies travel well but can't carry as much information. The higher ones don't travel far. This is one of the reasons why WiFi has been a lot faster than cellular services. The other is contention. The further your signal travels, the fewer access points you need (which is good), and the more people who are sharing the same signal (which is bad).
Some relatively new developments are likely to take some of the strain off this situation, although they are (as always) very processor-intensive. If you've watched analog television in a built-up or hilly area, you may have noticed ghosting. This is caused by one signal coming directly to you from the transmitter, and another bouncing off a hill and arriving slightly later.
In a built-up area, signals bounce off buildings, trees, and at some frequencies even people. For most existing protocols, you have to average these bounces to try to find the signal you really want. With more accurate equipment, you have another option. You can use the information about the different paths to identify a unique endpoint. Several devices can be transmitting simultaneously on the same frequency, in the same area, creating a lot more available spectrum.
While not entirely relevant, this diversion indicates that a fixed amount of spectrum doesn't mean a constant amount of digital bandwidth. Improvements in technology allow you to transmit more symbols in the same amount of signal space. But this fact doesn't alter the reality that, for now, spectrum is relatively scarce.
We currently have two extremes for spectrum allocated for Internet access: cellular and unregulated (WiFi and friends). The first is centrally controlled and covers wide areas. The latter is unregulated and anyone can use it, but it has only a short range.
Some of the technologies I mentioned earlier blur these lines a bit. A cable company could deploy a metropolitan area network quite efficiently by giving all of its customers WiFi access points and using a combination of mesh routing and Mobile IPv6 for hopping between them.
In the U.S., a decision was made recently to allow some spectrum formerly used for television (the so-called "white spaces" between the channels) to be used for data networks, too. This is in a nice range for metropolitan area networking, and provides another alternative.
The frequency ranges available here are even lower than those used for cellular services, and therefore the signal can carry a lot further (unsurprising, when you think how far apart mobile phone masts are, compared with TV transmitters). This makes it very attractive for providing wide coverage. Since the FCC decided to make this range available for use unlicensed, anyone can set up this kind of coverage. A town might have a lot of hotspots providing fast, short-range access, along with a larger base station covering the entire area with a slower connection as fallback. When you walk between two hotspots, you phone will fall back to using the white space system, without interrupting your coverage (although your available bandwidth may decrease).
In the future, expect to see your handheld devices jump transparently between networks controlled by different providers and offering different quality guarantees. Selling universal access, different levels of minimum connection quality, and a variety of other options will provide innovative companies with a range of areas in which to build new business models.