Most connections will be Wi-Fi or local. Very few require 5G high speeds or latency. Consumer IoT will come, but industrial is now leading the way. Those observations are from Linley Gwennap, one of the world's leading chip analysts. I listen closely to the chip guys, who often see trends several years ahead.
Wi-Fi is close to ubiquitous in every middle class home, so why pay a phone company to connect you in home or office? The vast majority of IoT apps are low speed and latency tolerant. Your air conditioner or washing machine doesn't need to communicate at megabits/second, much less gigabits. Adding Wi-Fi is just a couple of bucks and getting cheaper. 802.11ac is delivering a routine 500 megabits.
It falls off rapidly with walls, but beamforming and mesh will solve most of that. Linley believes, "Wi-Fi IoT devices will likely stick with 802.11n at 2.4GHz instead of adding the cost of a 5GHz radio for 11ac. IoT data rates are mostly less than 1Mbps, so 11n is plenty even when considering walls and such." To my surprise, Linley added Bluetooth to the conversation. "It offers 1Mbps, which is fine for most IoT devices. The problem with Bluetooth is the shorter range, but changes in the recent Bluetooth 5 release and new mesh versions should solve that problem, at least within the home.With VoLTE and VOIP now expanding, very little in most homes will need a telco wireless connection."
Connected cars, potentially the most important part of IoT, can connect locally (DSRC, 802.11p or ordinary Wi-Fi) as well as via today's LTE network. "Vehicle to vehicle communications needing very low latency can and generally will be served by direct connections." I hear from João Barros of Veniam, whose company supports hundreds of thousands of vehicles today. That may be the largest connected car network in the world. They have a great slogan, "The Internet of Moving Things." (Veniam is a very interesting company, that when I met them six months ago looked to be doing very well. One of their VC's is a friend. Ask me for an introduction if you'd be interested in their probably inevitable next funding round, but do diligence. I haven't seen the books.)
For ten next ten years or more, I'd expect most cars will do fine connecting with 4G. Remember, almost all networks are going to Gig LTE, with many times more capacity than we have today. What would you want to do in a car that needs more than 100 megabits? Gig LTE will have plenty of capacity for years.
Who would design a car or truck that wouldn't be run on most roads for 10 or 15 years? The thousands of semi-autonomous cars on the roads today are working well without 5G, using sensors, onboard computing, and existing wireless networks. 5G will not reach most parts of the world for at least another decade, I believe.
Remarkable things, I'm reporting, are happening in 5G millimeter wave, with NTT promising to do all of Japan by 2023. Verizon is planning a third of the U.S. by then, although they've said nothing directly.. China's MIIT has allocated $400B for next decade, although some of that will go to what is really 4G plus a software tweak. Japan and maybe Korea are the only countries likely to be fully covered within the decade. The Europeans almost universally don't see 5G as being profitable or necessary for many years. All will do a little to keep governments happy, but none have committed to more. Things are changing fast, however.
Linley points out that most IoT manufacturers are primarily interested in cost, not speed. He believes that really wide consumer use will require true ease of use, which he doesn't think many have yet achieved. On the other hand, there often a clear and prompt payoff in commercial IoT, so that's moving faster.
Almost everything using electricity will be connected, but I fear the different sectors are already expecting about 300% of the likely revenue.