5G: A framework for what's coming

Dave’s framework for understanding the first three years of 5G

• Networks take years to build.
• Technologies take years to move from the lab to wide deployments.

ergo: three-year forecasts should be reasonably accurate.

Surprises and mistakes are inevitable.

Update 10/16 Noted that indoor coverage is not as good as outdoor, but probably better than some believe; commented on less developed countries; clarified other parts.

Real (>90% likely within three years)

• Verizon will offer mmWave at speeds close to a gigabit to >10 million homes. It will spend most of a US$20 billion budget. It will not significantly raise capital spending. One of the first customer hookups tested at 1.3 gigabits at 2,000 feet. Then CEO Lowell McAdam said Verizon would be bringing a gigabit to almost all. The plan has apparently changed, to 300-1,000 megabits. They’ve provided no way to estimate how many would get the higher speeds. It almost certainly will be a more than a third of the 30M in the first phase.
• Verizon’s financial claims imply the cost of mmWave is much lower than most estimates.  France Telecom and China Mobile also claim the cost of 5G will be low enough to build with minimal capex increases. I believe they are right, despite the overload of claims that 5G will be very expensive. Lobbyists lie and regulators often believe them.
• mmWave can reach 65% of the urban area without needing new cells. (Qualcomm modeling) It will pass through some walls and windows but not all. The enormous capacity of mmWave will free the lower bands for better service where mmWave does not reach.
• Simply upgrading existing cells is a logical - and relatively cheap - way to deliver mmWave. This inexpensive build probably will cover 20-40%, although there is no hard data. CEO Hans Vestberg pointed out that prices always go down for telecom gear. He expects the price will come down enough that the second phase, in three or four years, will continue to fit in the same capex budget.
• Few if any other carriers will do a large deployment of mmWave. As I write in October, 2018, John Donovan says AT&T is holding back after initial mmWave deployments. He has upped the pace of FTTH from 3M/year to 5M/year. No others have announced mmWave beyond trials.
• Nearly all 4G being installed today is designed for a 5G upgrade. Ericsson tells me that just about all they are shipping since 2015. Most will be upgraded to “5G NR” fairly soon. By 2023 or so, a large percentage of radios will be upgraded. Therefore, I expect wide coverage of “5G” fairly soon.
• Most carriers will do a fairly large deployment of mid-band called 5G. Mid-band 5G is 4G hardware with NR software, comparatively inexpensive. The software may raise the capacity by 15% to 50%. (Similar spectrum and antennas.) That’s less than the improvement from adding LAA to LTE. Typical downloads should be 100-400 megabits.
• Many carriers in the developing world will offer mid-band 5G. I expect volume in 2020 to 2022. India is likely to be the leader. Reliance Jio has won 225 million 4G customers in just two years. I believe most of their gear is from Samsung and upgradable.
• The available capacity gain from 4G technologies is six to twelve times most places, without using either millimeter wave or New Radio. That is enough to meet the anticipated growth most places well into next decade. Growth is slowing. Cisco expects 30% in the U.S. in 2021. Article to follow.
• Whether called 4G or 5G, I expect strong sales in the developing world. The boost from Massive MIMO and Carrier Aggregation is crucial in Africa, India, and anywhere there are few wires. Affordable offerings today are often 5-10 gigabytes/month. That’s not enough for a video Internet. People want and deserve five and ten times that. Telcos will be smart to meet that demand. Making it 5G by adding NR software adds only modestly to capacity (15% to 50%.) I don’t know how quickly the software will be added. I expect the software price will be more than recovered by the marketing advantages in the developed world.
• U.S. cable companies will go big into wireless. “Humongous” says a senior tech person. Cable people are swarming 5G discussions. Cable has fibre and power close to most homes, bringing the deployment cost all the way down. John Chapman of Cisco is considering picocells in every home gateway for even more capacity. Unless the telcos offer a remarkable MVNO deal, this will be big.
• Carriers will say their 5G is wonderful and even revolutionary.
• All new networks will claim major achievements in software-defined networks, artificial intelligence, and similar. Most will find progress slower than hoped.
• CTOs will continue under enormous pressure to cut costs. Verizon has just offered buyouts to 44,000, a quarter of employees. Vendors realize this. Huawei has changed its messaging to how it drives down costs rather than the capabilities of the equipment.
• Politicians will claim everything good is due to their wise policies. Rarely will any policies have much impact on what is built, unless a government makes strong demands. The common “incentives” are more often giveaways and rarely inspire much additional investment. (Policy does have an effect on prices and profits. Investment is driven far more by demand, technology changes, and competition in most cases.)
• China will be far ahead in the number of cells except in mmWave. China mobile has two million cells in their plans, most at 3.5 GHz
• 5G phones will be expensive. They will be unable to support most of the expanding number of bands.
• Apple will support 5G, highly limited in 2019.
• Huawei will be the largest supplier of telecom equipment. Huawei’s US$15 billion annual investment in research and development all but guarantees a top spot.
• Cost per bit delivered will go down. Verizon estimates 40%/year.
• Price consumers pay per bit will go down. In competitive markets, the fall will be similar to the cost drop. Carriers will always seek to signal competitors and keep prices higher. They call that “rational pricing.” In the U.S. in October 2018, Sprint and T-Mobile are trying to get higher prices in absolute terms in the U.S.
• Industry engineers and clued in executives will generally know what is going on. Most others will not.
• There will be talk about 6G. Already, I am on the ITU Focus Group 2030 setting the parameters. It’s open to all; ask me how to join.
• Verizon will sell a meaningful number of fixed wireless connections.
• Surprises will happen and there will be some errors in this.

Probable

• With volume, the difference in price between "4G" and "5G mid-band” will be small. The hardware of 4G and 5G today are identical. The low cost is why “5G” will have huge gains 2021-2024, becoming near-ubiquitous in many countries. (Older 4G radios will need a hardware upgrade.)
• Verizon's fixed wireless will sell better than expected by most. Goldman Sachs estimates 8 million by 2023
• Capital spending and hence overall vendor sales will be close to flat. Verizon is building mmWave without raising capex. Mid-band 5G is 4G hardware with a software tweak and costs the same as 4G.
• Deutsche Telekom and U.S. telcos in D.C. will continue putting out wildly exaggerated cost projections. They are looking for government favors and subsidies. To the extent these have any basis in reality, they reflect a network that will never be built, millimeter wave every 200 meters along every small road with a large server included. (Edge networking to the extreme.) Politicians inclined towards the telcos will believe even absurd claims. This need not be corruption. Confirmation bias - our mind’s strong inclination to discard anything that that would change our beliefs - is very strong in the political class.  
• Qualcomm will be the leader in chips. Samsung and Huawei’s HiSilicon will make most of the chips for their companies. Intel and MediaTek are currently behind but hope to play a major role. Spreadtrum/Unisoc, in close collaboration with Intel, also promises a chipset in 2019.
• Sales of fixed wireless between 50 and 200 megabits will be strong. Almost all wireless carriers will offer the service, but the take rate is unproven.
• Very modest deployment of Edge networks in Europe and the U.S.

Unknown

• Rate/reach of millimeter wave. This is huge because of the effect on the cost of rolling out service.
• The take rate of the higher speeds. On the wireline side, many are buying the gigabit even if they have little practical use for it. If the highest speeds are reasonably priced, I expect the same on wireless but that’s unproven.
• The take rate of 5G at mid-band speeds. 4G LTE speeds can be similar, but my guess is the “5G” label will increase sales.
• Effect on consumer prices. Prices are driven mostly by competition rather than the cost of service.
• Effect on company profits
• The relative position of suppliers after Huawei. Ericsson has a good chance of remaining close. Samsung will see fast growth. How many customers will choose to buy 5G phones when 4G phones meet almost all practical needs.

Unreal (<10% likelihood in three years.)

• 1 millisecond latency beyond trials. No carrier has announced deployment plans. Many are skeptical it will ever find a large market.
• “5G” much faster than good 4G unless mmWave. T-Mobile LTE/LAA is at 500 megabits here in Manhattan. BT delivered 750 megabits to Wembley in 2017 using 4G.
• 5G upgrades are required to meet the traffic demand. Traffic growth is slowing. 4G can increase capacity 6X-12X, using Massive MIMO, carrier aggregation to put more spectrum to use, 3.5 GHz, LAA, and an economic number of small cells. Verizon thinks they will find services that will need the mmWave capacity, but haven’t suggested any. They intend to use the enormous capacity for a lasting competitive advantage.
• Necessary for IoT. IoT works fine on 4G & Wi-Fi
• Necessary for telemedicine. Almost all telemedicine features work fine at ten megabits, much less 200 megabits. Unless you believe most surgeons will operate from the beach, remote surgery is not dependent on 1 ms wireless. Nearly all remote surgery will be conducted from indoors with landlines available.
• Needed for autonomous cars. Autonomous cars must be reliable without coverage, hence must be designed not to depend on it.
• A big economic boost. All those claims are b______. The 200 megabits of LTE are enough for almost all volume applications, so 5G will not induce economic growth. LTE latency can go down to ten and even two ms.

Uncertainty becomes much higher in four to six years. Initiatives begun now have more impact. Outliers, like Verizon’s millimeter wave build, may prove successful and inspire others to change their plans.

This is moving extremely rapidly. Competitive strategies may emerge. Vodafone and Telefonica in Germany are considering disrupting the market with mmWave. The technical people like the idea, but the CEO and CFO do not agree today. If either moves, others will follow.

We have no verifiable data on take rate and actual costs. Without data, I can't be accurate and the numbers four and five years out are very fuzzy. Clients demand analysts provide five-year numbers to fill in their spreadsheets. Rarely are long-term forecasts very accurate.

Overall, I expect most of the above to hold true through 2024-2025. But here are some possibilities to watch:

• Verizon’s mmWave may be successful and inspire others to follow. The first phase of a mmWave buildout can feature upgrades to existing cells, keeping the cost in the first few years surprisingly low. My guess is that the best strategy for the next few years is to upgrade most existing cells to mmWave, but only Verizon has announced plans to do so.
• Some carriers may look for a competitive advantage by doing mmWave, forcing others to match. (Makes sense in Germany and Canada.)
• The premium for 5G phones will shrink. Moore’s Law isn’t dead but is not as rapid. The cost of 5G chips may take a while to fall.
• Around the world, carriers will do more sharing of networks. The Chinese share a tower company. Bell Canada & Telus share a network. One network is cheaper than two and two networks are cheaper than four.
• Lower cost equipment will extend advanced networks in rural areas and less developed countries. Huawei’s RuralStar is aiming at that market.
• As wireless networks get better, more people are cutting the cord for data. So far, the number has been modest. It will grow, possibly at a fast rate.
• In some places, demand will reach the economic limits of sub 6 GHz and mmWave the necessary choice.
• Cable companies have fibre and power close to homes. That enables them to go inexpensively into wireless. Some will be very successful.
• Edge networks will grow, possibly dramatically. China will lead the way. Minister Miao Wei expects “90% of China within 25 ms of a server by 2025.”

(Better ideas welcome.)