"There is a high degree of uncertainty," Ovum's Mike Roberts correctly notes in his estimate, but his 24M projection is consistent with other datapoints. (My wild guess would be lower.) MU-MIMO and Massive MIMO - in sub 6 GHz frequencies - will almost certainly be more important into the next decade.
High frequency millimeter wave will be an important part of 5G many places one day and working on rules is sensible. The hype, from the U.S. FCC and the vendors, is so inaccurate it's almost funny.
The most aggressive promoters of millimeter wave - Verizon, NTT, Korea Telecom - have been clear: trials only until 2020-2024. VZ CFO Fran Shammo made a point of telling Wall Street that the actual deployment will be so small it won't affect capital spending for years. NTT CTO Seizo Onoe in 2015 said he expected his 5G deployments mostly will be below 6 GHz until 2022-2023 and confirmed that this year. Giant China Mobile just joined Qualcomm in a demonstration of 5G in sub-6GHz spectrum at Mobile World Congress. (Release below.)
The FCC's current effort to allocate 28 GHz and other bands for 5G is not a mistake, however. Verizon intends to put up a trial system within eighteen months. That's mostly for PR, but VZ is very serious about using 28 GHz to replace the 10 million lines of copper they want to shut down in the next decade. The technology will work, but the number of base stations and the related costs still need to be determined. The U.S. eff you attitude to WRC international standards is inappropriate.
High frequency 5G is a competition killer, at least as proposed by AT&T and Verizon. Unless near-monopolies are acceptable, very strong measures will be required to allow meaningful competition. Both Democrat Jessica Rosenworcel and Republican Mike O'Rielly have been promoting 28 GHz very aggressively while burying in the sand about steps that would maintain competition.
Competition on telco local loops is technologically obsolete. England could raise Internet speeds 100 megabits if they understood the old form of unbundling will fail. It may be that the cost of millimeter wave most places will make it a natural monopoly or duopoly. Ten times more base stations needing ten times more backhaul will be too expensive for most carriers. Building the 4-7 networks usually needed for competition to work well is almost impossible. Existing telcos and cablecos - with lots of fiber in place - have an enormous advantage.
"It's an incumbent's game," a major CTO tells me.
The sleeper here is the 7 GHz of unlicensed spectrum around 60 GHz WiGig. That's coming in a few years in mobile phone chips and could be game changing.