Cell bandwidth is also approaching full load. It's going to take a fundamental redesign of frequency management (which looks promising) before bandwidth gets cheap again. And when that happens, I'm a tad skeptical that monthly fees will drop-- until the entrepreneurs start undercutting the global megacorps.
Starting in the early 90s, when I worked with a start-up in this field, frequency rights were auctioned off by the FCC to telecoms who paid billions of dollars for the rights to very specific markets. Then, they had to build cell towers, which cost $500K each then. Recently, I was told the price was up to a few $million a tower. So, wireless internet has to be expensive.
Wireless links will always have a total capacity limited by the link bandwidth. Wired links are subject to bandwidth limit too for that matter, but one can easily add more wire, while broadcast spectrum, like land, cannot be created. Different modulation schemes are just ways of getting closer and closer to that theoretical Shannon limit. I am no longer working in that field (stopped not long after the debut of the 1G), so do not know where they are now, and how much room is left to exploit.
Another way of dealing with increasing traffic is by making a cell area smaller. Of course this costs more money for more microcell towers. It seems like more and more people are using their smart-phone for apps that used to be on the PCs, which get internet connection via land lines meaning cable modems, DSL, ISDN, T1 etc... Expanding these seems a lot less expensive, and indeed the amount of traffic that my household pulls through the cable modem would bankrupt me, if I were to pay wireless rate for it.
Anyway, the Navy guys should be happy with semaphores. What is its equivalent baud rate, I wonder? It could not be too far from the baud rate of their VLF link, right? And both, I am sure, are high speed communication relative to the smoke signal that the Indians used. Heck, wasn't that about 1 bit per 10 seconds? What would be the highest bit rate of a proficient Indian smoke signaler, under ideal conditions (good visibility, no wind, etc...)? And I mean an error-free bit rate, not that of an ambiguous message that would have to be retransmit due to lack of confirmation.
PS. Just had an idea. Has anyone thought of augmenting smoke signals and semaphores with extra signals encoded for parity error detection? And then, how about more advanced forward error correcting codes? Oooh, could be an area for research and then market development here.
PPS. My memory failed me. It was the ELF link that could really reach submarine depth. It has a speed of a few bits per minute! Now, that is slower than smoke signalling, I think.