To get decent coverage you need to negotiate and pay for space, power, and internet connectivity on thousands of towers/tall buildings all owned by different people and municipalities who see you as somewhere between an annoyance and a revenue stream to exploit. Then you need a field workforce capable of performing maintenance at all of those sites (which requires making appointments and sometimes paying fees to the property owners), a supply chain of spare parts, and the RF expertise and test equipment to measure and do quality control on their coverage all around the cities you're operating in.

Then you need a way to recoup all this cost.

Being a cell carrier takes staggering amounts of money and staggering amounts of schlep. It's not for hobbyists, hackers, or small companies (who are not in fact carriers but just resell and rebrand real carriers' services). There is a reason it's the domain of giant corporations run by the kind of people who make deals (and not write code) for a living.

This is one of the reasons its a fun idea. The TV bands (50 - 210Mhz) can cover a huge area from a single antenna (that was their original modality) but because of the inverse square law, it means that a much smaller and lower power antenna nearby can still get a big chunk of real estate. During some experiments at a company that will remain nameless it was something like 1 antenna could easily cover 2 mile diameter circle, 4 would cover 8 square miles, and 30 - 40 could cover nearly all of the cities around the SF bay.

The actual cellular mechanics are quite an undertaking, but something along the lines of a coded point to point system would be implementable by a small group of people.

Again, the problem is getting the antennae up high. Unless you and your small group happen to own some radio towers and/or tall buildings, range is going to be much much smaller.

Running a cell provider is not that hard.

A lot of companies have their own towers that only provide service within of a city, and for events, you can even set up your own network, like the CCC always does during their conferences.

Having a national, or even global network is a lot harder, though.

You're better off assuming the network is hostile and then encrypting at a higher layer, ZRTP, OTR and so on.

+ for ZRTP and OTR - good comment. Always assume the network is compromised.