We have only guesses for now, but $tens of billions of decisions are being made. Until we have extensive data from the field, everyone is relying on limited lab trials and models. There are no proven facts about mmWave economics. I can't do informed reporting without some numbers to use, so I'm collecting all the data I can find. Many answers also require estimates of traffic demand, how much people are willing to pay for what technology, whether connected cars require 1 ms, 5 ms, or 10-30 ms latency, and many some unknown unknowns. Improvements necessary and welcome.
Many answers also require estimates of traffic demand, how much people are willing to pay for what technology, whether connected cars require 1 ms, 5 ms, or 10-30 ms latency, and many other datapoints not covered here. In particular, "unlimited" offerings are driving up traffic demand far above expectations even six months ago.
Can Verizon make money selling fixed wireless? Will sales of 5G gear turn around the declining business at Ericsson & Nokia? Was Google right to resume installing fiber in Nashville? Will m mmWave be able to serve the 50+% of "mobile" traffic from indoors? Will rural areas ever be financially practical? Will competition die because 4-7 just can't be supported?
How many homes actually served per transmitter?
Nokia estimates "each 5G base station can reach tens of households, each fitted with an antenna," with a share of 5 gigabits download. That would deliver a gigabit to most of the homes nearly all the time. Others make much higher estimates for how many you can serve with that much bandwidth to share.
How many homes can be covered per transmitter?
Chet Kanoja at Starry expects 1,000-2,000 homes with a triple beam transmitter in an urban area like Boston. He thinks his beams can bounce or bend. Most other estimates are between 150 and 400 homes. Verizon is hinting they will do better than that but haven't released data.
mmWave can go 1,500 meters with good line of sight but not even 6 inches if certain walls or windows are in the way. It definitely can't get to the other side of a building. Depending on building layout and density, the number served could be anywhere from less than 50 to more than 1,000. Most analysts are thinking 200-300 meters.
How much will it cost?
The cells are coming in cheaper than people expected, but I can't find any reliable numbers. Starry uses a ptmp microwave, a much simpler and cheaper unit than LTE. Kanoja told Craig Moffett "each 'Starry Beam' costs $7,000 to $8000." That's way more than the $1,000 BOM Kanoja quoted me last fall but not far out of line with some commercial five gigabit point-multipoint mmWave systems.
An LTE system like Verizon plans will be much more complicated and expensive, but Verizon will ask for a bid in quantity 250,000. The vendors really want this contract and Verizon has brought Samsung in as a possible alternative to Nokia & Ericsson.
Kanoja also is seeing a $1,000 cost per receiver, far from economical. mmWave receivers are not a volume product; the cost will come down. It will also cost about $200 to install the outside antenna and connect the home. With advertising and overhead, that's a customer acquisition cost of $500-$1,000
Nokia calculates a 5G mmWave home offering would have to charge $43 minimum. When I add in customer acquisition costs and telco expected margins, I think that number will be $60-$70, similar to Google's gigabit fiber. Kanoja is a good engineer with $63M in venture capital. Even so, he'll have to bring costs way down or take rate way up to be profitable.
The cost of the transmitter and opex has to be spread over the numbers of homes taking the service. If a $7,000 transmitter can pass 350 homes, that's $20/home, If 10% take the service, that's $200 for the transmitter. The Starry and Nokia assumptions seem to imply a higher take rate than I think likely.
I believe most customers are happy with the 50-100 megabit cable service.The evidence around the world is customers do not want to switch, reluctant to have to learn the quirks of a new system, possibly reconfigure the home Wi-Fi, lose time from work for the install, and master a new customer support system. France Telecom/Orange discovered that even customers with slow DSL did not want to switch to 100 megabit fiber home when it became available. They had to actually drop the price of fiber to below DSL prices. Google ran a massive marketing campaign, offered a symmetrical gigabit with generous terms, had some of the most admiring press in the world, and found most people in Kansas City still didn't sign up.
(Verizon expects to soon add mobile 5G for more direct income. They expect they will win customers and sway policy with the advanced image of the first in the world. Their situation is much more promising. I believe they have decided to go ahead with a $20B network for 1/3 to 1/2 of the U.S. Nothing is likely to be announced for a while, however.)
How much will the phones cost and how good will they be?
Linley Gwennap, a top engineer estimates mmWave phone chips will require 10x as many transistors as LTE phones. At least the early ones will be hot and kill the battery. The necessary components will be hard to fit into the standard cell phone size. The 2019-2021 models will be at or beyond the state of the art for chips. The complexity of many phones moving around and supporting high data rates shouldn't be underestimated. Getting something decent around 2020 would represent extraordinary engineering. Qualcomm and Intel are the best in the world; I hope they deliver.
Just like the 2017 Gig LTE phones, they will only be affordable for the high end for several years. I'd expect a large subsidy will be needed, but it's really too early to guess.
Few of us need a gigabit to our mobile, or even 100 megabits. Sales volumes may be low for years.
(This is wildly speculative until we have far better data.)
How fast will it be?
Huawei and others have demonstrated 20 gigabit downloads but that will be rare in the first generation. Verizon suggests 5 gigabits in early days. 5 Gigabits split between dozens and even hundreds of users will deliver a gigabit to most homes nearly all of the time. Very few homes will be pulling regularly more than required for 2 HD video streams, about 10 megabits. While it's possible seven users will each seek 800 megabits at the same time, that's almost never seen in practice.
When we get to phones in motion, the speed will probably drop. Gigabit LTE seems likely to deliver hundreds of megabits most of the time; 5G will certainly be faster, but we have no meaningful data yet.
What latency can we expect?
Brave talk of 1 ms latency doesn't seem to be realistic. Nokia - which elsewhere promotes 1 ms - defines extremely low latency "below 5ms." AT&T projects 5 ms and I predict Verizon will be similar. Both are using Cloud RAN to control multiple cells from a central location. Getting to and from the C-RAN requires several hops through switches and routers. I haven't seen anyone promising C-RAN latency below about 5 ms.
It's possible to go from a cell a few hundred meters in a millisecond, but a network like that requires expensive processing at the edge. That drives up the likely cost so much, no major network has promised to support 1 ms. For a network the size of Verizon's or AT&T's, the extra cost would be in the $billions. Deutsche Telekom suggests the cost across Europe might be 50 billion euro.
How many customers will buy the service?
As noted above, I'm not expecting rapid takeup. I'd be delighted if I prove wrong, of course.
With high fixed costs to build the network, a decent take rate is required for any profit. Verizon's prior CFO, Fran Shammo, was also skeptical.
Nokia's flowchart, below and at top left, is thoughtful and logical. Unfortunately, they held back essential figures, which means I can't support their conclusions.