China is just getting started installing what will very soon be over a million 5G cells. They will use frequencies from 2.5 to 4.9 GHz and deliver typical speeds of 100-400 megabits. What apparently is undecided is whether they will aim for ~30 ms (Verizon), 15-25 ms (Deutsche Telecom), or 10-15 ms? 

30 ms is the least expensive, achievable by replacing older LTE with brand new 5G NR. That saves 35% to 45% from older systems most common in the U.S. That can be reduced by to 15-25 ms by installing edge servers and/or reducing the number of routers. (Verizon intends both.)

No one in the west has announced plans to go to every tower, which would require about 10X the number of servers. Highly respected Minister Miao Wei and Xi Jinping have hinted China might do so, and later cut the latency in half when URLLC is ready.  Nothing is clear, especially because I have to read in Google translation.

Xi Jinping sent a message to an AR/VR conference, "Virtual reality technology is gradually maturing, expanding human perception and changing product forms and service models." AR/VR is the only major application today that requires low latency. The general opinion is that anything over 20 ms can cause nausea and 10 ms is better. Verizon's current 5G at 30 ms is considered inadequate for AR/VR. 

Minister Miao in March said, "5G technology applications represented by driverless cars will become an early application. The world is working to advance the development of driverless cars.” Experts in the West generally have concluded, ".'" Cars can't freeze when they go out of the range of the network, so they have to be designed to work without it.

Miao in 2016 said his goal was 25 ms in 2025. The technology, especially of Edge Networks, has improved since then. 15 ms is now economically practical and 10-12 ms not impossible.  5-10 ms will be possible when URLLC comes out of the labs although that isn't close. 

If in fact the car is controlled by the network, these lower latencies will probably be necessary. If the main use of connected cars will be information and entertainment services, very low latency 5G is probably not needed. Those services can be served by 4G or slower 5G.

Those indirect comments are provocative but not enough to make a clear prediction.

 

Carriers have told investors most of their 5G plans for the next few years, allowing a fairly accurate prediction of what will be in most major countries.  Most will deliver customer speeds of 100-400 megabits to most, with a latency of 15-30 ms. Less than 15% of the world will get millimetre wave, with speeds typically three times higher and often a gigabit.

 

dave ask

Newsfeed

The 3.3-4.2 spectrum should be shared, not exclusively used by one company, concludes an important U.S. Defense Innovation Board report. If more wireless broadband is important, sharing is of course right because shared networks can yield far more

It does work! Verizon's mmWave tests over a gigabit in the real world. 
The $669 OnePlus 7 Pro outclasses the best Apples and probably the new Galaxy 10 or Huawei P30 Pro. Optical zoom, three cameras, liquid cooling, Qualcomm 855 and more.
Korea at 400,000 5G May 15. Chinese "pre-commercial" signing customers, 60,000-120,000 base stations in 2019, million+ remarkable soon. 
5G phones Huawei Mate 20, Samsung Galaxy 10, ZTE Nubia, LG V50, and OPPO are all on sale at China Unicom. All cost US$1,000 to 1,500 before subsidy. Xiaomi promises US$600.
Natural monopoly? Vodafone & Telecom Italia to share 5G, invite all other companies to join.
Huawei predicts 5G phones for US$200 in 2021, $300 even earlier
NY Times says "5G is dangerous" is a Russian plot. Really.
Althiostar raised US$114 million for a virtual RAN system in the cloud. Rakuten, Japan's new #4, is using it and invested.
Ireland is proposing a US$3 billion subsidy for rural fibre that will be much too expensive. Politics.
Telefonica Brazil has 9M FTTH homes passed and will add 6M more within two years. Adjusted for population, that's more than the U.S. The CEO publicly urged other carriers to raise prices together.
CableLabs and Cisco have developed Low Latency XHaul (LLX) with 5-15 ms latency for 5G backhaul,  U.S. cable is soon to come in very strong in wireless. Details 
Korea Telecom won 100,000 5G customers in the first month. SK & LG added 150,000 more. KT has 37,500 cells. planning 90% of the country by yearend. 
The Chinese giants expect 60,000 to 90,000 5G cells by the end of 2019.
China Telecom's Yang Xin warns, "Real large-scale deployment of operators' edge computing may be after 2021." Customers are hard to find.
Reliance Jio registered 97.5% 4G availability across India in Open Signal testing. Best in world.

More newsfeed

----------

Welcome On Oct 1, 2019 Verizon turned on the first $20B 5G mmWave network with extraordinary hopes. The actual early results have been dismal. Good engineers tell me that will change. Meanwhile, the hype is unreal. Time for reporting closer to the truth.

The estimates you hear about 5G costs are wildly exaggerated. Verizon is building the most advanced wireless network while reducing capex. Deutsche Telekom and Orange/France Telecom also confirm they won't raise capex.

Massive MIMO in either 4G or "5G" can increase capacity 3X to 7X, including putting 2.3 GHz to 4.2 GHz to use. Carrier Aggregation, 256 QAM, and other tools double and triple that. Verizon sees cost/bit dropping 40% per year.

Cisco & others see traffic growth slowing to 30%/year or less.  I infer overcapacity almost everywhere.  

Believe it or not, 80% of 5G (mid-band) for several years will be slower than good 4G, which is more developed.