The gleam in Lowell McAdam's eye discussing his future network was a hint. Verizon's purchase of $1B of Corning fiber convinced me the decision had been made to go to a one-third to one-half the United States. I have no inside information; the CFO didn't have too many drinks one night and spill secrets. Rather, the evidence and the logic drove me to believe this will soon be the largest network build in the Western world. It really doesn't look like Verizon will buy Charter (or whomever.) Making looks much cheaper than buying. It's also possible the 2017-2018 testing will be severely disappointing. They can change their decision until 2019-2020, when the spending gets heavy.
The evidence below needs to be considered alongside this separate article about what's inspiring Verizon to go ahead. The technology is (almost) ready. 5-20 gigabits is working in the labs and 10,000 engineers around the world are pushing forward. VZ is being killed by cable across the unfibered half of their territory. AT&T is making noise about going out of territory with wireless to the rooftops & gigabit to the apartments. The business logic is strong.
Verizon decisionmakers can see what I'm seeing.
Ivan Seidenberg left a legacy of two great networks: Fios & Verizon's early 4G to 98% of the U.S. In both cases, the technology was unproven but he took the risk. Both were delivered on time and on schedule and changed the industry worldwide.
The world's first large 5G millimeter wave network would be the legacy of Lowell McAdam.
Most of the cost of a fiber network is labor on the ground, not the fiber optic cable itself. They announced they are buying over $1B of cable from Corning. That implies the total spend will be $4B-$7B. They just spent $1.8B buying a 20,000 mile fiber network from XO.
It will save money on backhaul and win some more customers. That isn't enough to justify the expense unless they also are building an enormous small cell network.
They would be stupid to do this unless they needed massive fiber for small cells backhaul . Verizon's top executives aren't stupid.
Auction prices were 50-70% cheaper this year than the previous auction, in which they spent $10B. They have 40 MHz fallow, enough to about double current capacity. Going to 4x4 MIMO and 256 QAM will double it again. But traffic growth is accelerating since U.S. carriers went unlimited. Growth was down to 40%-50% and Cisco predicted it would go below 40%. We only have a few months of data, but that trend is reversing. A competitor claims Verizon network speeds fell 14% while others were increasing.
Somewhere between 2020 & 2023, they will need a way to keep up with the traffic. If they weren't confident about 5G coming quickly, they almost surely would have bought some spectrum. Meanwhile, the small cells can run "Gig LTE" or similar and add the needed capacity.
Seizo Onoe, CTO of giant NTT DOCOMO, last year predicted we wouldn't see much 5G millimeter wave until 2022-2023. Among top engineers, Ted Rappaport was virtually the only one publicly predicting sooner. At the Brooklyn 5G, Onoe startled everyone by saying things have changed. He now predicts substantial builds coming in 2020 or 2021, probably including mobile.
At Brooklyn, both AT&T & Verizon promised "commercial deployment" of 5G fixed in 2019. Verizon believes they will win a significant number of customers from cable for 5G fixed. Likely speeds are 5 gigabits shared, a realistic 1 gig to most homes most of the time. Cable is now routinely 50 meg down and higher, enough for most people. (I have 200/20 for only $15/month more.) While Comcast intends to do some DOCSIS 3.1 upstream in 2017, most cable companies are waiting for Full Duplex for fast upstream. They expect it by 2019-2020, but I haven't seen even a demonstration unit. Verizon believes they can beat cable with the much faster upstream. If people are reluctant to switch, customer acquisition costs may be so high 5G fixed may not prove profitable. (Unproven - no data.)
Intel and Qualcomm are racing to produce the first chips for 5G phones, no later than 2019. The first will probably be large, run hot, and drain the batteries. But Moore's law isn't dead for at least two more generations. Intel, Samsung, and Qualcomm's main foundry, TSMC, are just now getting to "10 nanometer" production, All have announced plans fo 7 nm and 5 nm generations on a fast schedule.
Earlier deliveries significantly reduce the time to revenue of any 5G build.
Verizon and AT&T are accelerating their trials.
A few months ago, Verizon was talking about 2017 testing to a handful of homes near their labs in Boston. Suddenly, they are coming to eleven cities across the country by June. They want to quickly develop data about results in different terrain. mmWaves have trouble penetrating many windows and walls. They won't go through buildings. Some believe beamforming will provide decent performance without line of sight. Whether and when that applies is totally unknown until we have field data.
5G is coming in cheaper than expected even a year ago.
Unconfirmed numbers are as low as $100/home passed. Nokia & Ericsson, the prime U.S. suppliers, are facing Samsung, with Wonil Roh and other respected engineers. Meanwhile, both Ericsson and Nokia need to win these contacts. Ericsson predicts the worldwide market for telecom gear is going down ~6% this year. I believe Huawei is winning a greater share. The U.S. and Australia are the only major markets Huawei is not a factor.
There's no public data on pricing. Starry's Chet Kanoja, a competent engineer, believes he can bring in a one gig mmWave transmitter for a parts cost of ~$1,000. (He uses off the shelf 802.11ac chips and changes the frequency.)
Verizon is moving quickly to Cloud Ran, much cheaper intelligence for multi-cell deployments. C-RAN is proving far more flexible and easy to update. No one heeds to climb towers. C-RAN almost certainly cannot deliver the much hyped 1 millisecond latency; AT&T is hoping to get down to 5 ms. The multiple routers/switches inevitably slow things down. Few telcos believe there's a large enough market for 1 ms to justify putting all the cost at the edge. Even most connected cars don't need that speed. I have not heard any telco, anywhere, commit to a large 1 ms network. Deutsche Telekom and friends are demanding a $50B subsidy for mmWave with 1 ms along all roads.
5G "New Radio" is a severely watered down system that is mostly 4G controlled. It will get to market faster and cheaper.
3GPP has developed a hybrid 4G/5G radio standard that should be much cheaper to produce. Many "5G" features are stripped out and 4G systems will provide the control plane. Millimeter way will be an auxiliary data channel, adding capacity when required. "NR" does implement some coding and other 5G features, but it's mostly 4G with extra data capacity.
This was demanded by a worldwide group of telcos determined to have something called "5G" sooner, mostly for pr and lobbying.
LTE, almost everyone now agrees, will not be replaced by 5G many places for at least 10-15 years.
Beamforming apparently is making a major difference in mmWave performance. (Unproven.)
Folks like Starry are claiming real-world mmWave reach of 500-1,000 meters. Verizon seems encouraged but has said little. If true, this can raise the number of homes reached on average from 150-300 to as many as 1,000. Absolutely no field data and little more than hints from lab testing.