Some network capex assumptions seem remarkably stable after 50 years. An example is the adage that a fixed network operator “has to do something” for about 10 percent of the physical plant every year.
That has proven a useful rule of thumb for cable TV and telco access networks, and now also seems to be useful for mobile networks.
Essentially, that means every fixed or mobile network is potentially 100-percent renewed every decade.
Up to this point, mobile cost per gigabyte has been an order of magnitude more costly than fixed network cost per gigabyte. That has meant that mobile internet access is not generally a full substitute for fixed network services, where both mobile and fixed internet services are ubiquitous.
The big change with 5G is that cost parity might be possible in some segments of the market, and perhaps as much as half the market or so. In the United States internet access market, for example, most fixed network customers buy service plans in the 100 Mbps to 200 Mbps range, for recurring prices in the neighborhood of $40 to $50 a month.
So the immediate target for a would-be competitor in the home broadband market is to offer speeds and costs in line with those figures, which allows mobile operators to compete in perhaps half the home broadband market.
The importance of 5G is that, for several reasons, the cost of supplying a gigabyte of usage will drop, compared to 4G, as the cost of each successive mobile generation has done. Cost per gigabyte has been steadily declining, for both fixed and mobile networks, for decades.
As always, it matters how we count.
The posted retail prices are not necessarily the “actual prices” consumers pay, as many are on promotional deals at any particular time. The other issue is prices for actually-used capacity versus plan allowance price. They are usually different.
The nominal (designed-for rate) is total usage allowance divided by total recurring cost. But not many users actually consume all the data their plans provide. Also, customers on unlimited-usage plans will have highly-variable “cost per consumption” ratios, as price is fixed, while usage is unlimited.
Fixed network data costs, on a cost-per-megabyte basis, routinely have in the past been in the 20 times to 60 times lower scale than mobile data. Where fixed network data might cost cents per gigabyte, mobile data costs dollars per gigabyte, counting either plan costs or actual usage costs.
Two key matters are the retail price of a service plan, which is shaped, in turn, by the cost of supplying the capacity, usually denoted as “cost per gigabyte of usage.”
Determining price on a country-by-country basis also requires adjusting prices to account for differences in currency values and purchasing power (typically using a purchasing power parity method). All those things done, price per gigabyte of mobile data usage ranges between nine cents per gigabyte up to $110 per gigabyte, according to Speedcheck.
There are all sorts of other complications, including the speed of a connection; whether we include fees and taxes as part of the calculator and whether other changes, such as equipment rental, also are included. All of that will differ from country to country and provider to provider.
But a reasonable rule of thumb has been that mobile data costs an order of magnitude more than fixed network data, on a cost-per-gigabyte basis. So in the U.S. market fixed network gigabytes might cost 30 cents while mobile gigabytes cost $3.
If mobile bandwidth traditionally has been an order of magnitude more expensive than fixed network bandwidth, then it is obvious that, to compete, mobile bandwidth has to be as capacious and affordable as fixed network bandwidth.
What is clear is that, compared to past capabilities, 5G networks will have a cost-per-gigabyte profile that allows mobile operators to radically close the cost gap with fixed networks that prevailed with 4G and prior mobile generations.
Using mobile networks to compete in the home broadband market never gets the headlines when we talk about 5G. The buzz is all about edge computing or internet of things or virtual reality.
We might be surprised by the near term revenue upside. Mobile operators might make more new revenue from home broadband services--however unheralded--than from edge computing, IoT or AI-based apps.
How much do U.S. mobile operators spend each year on capital investment? It depends on what we count.
Definitions and assumptions always matter when evaluating such matters as mobile industry capital investment. Among the most consequential matters is whether money spent to acquire spectrum licenses is counted as “capex” or not.
We might use standard accounting rules or industry conventions, but we need to know which definition is being used, as it makes a big difference.
Consider a new CTIA report on U.S. mobile industry capital investment. Between 2016 and 2020, for example, “capex” is said to range from $25 billion to $30 billion per year.
Spending for spectrum licenses ranges is tracked separately, it appears, at least in part because, historically, spectrum license spending has been highly variable. Prior to 1994, spectrum license spending might have been tracked by decade. Between 1994 and 2001 CTIA uses an eight-year interval. Since 2002 CTIA uses a five-year interval.
That indicates a quickening of the tempo of license acquisitions; more competition for licenses as well as bigger spectrum allotments (more frequent auctions; more licensees; more capacity per auction).
It appears CTIA, for purposes of tracking capital investment, does not include spectrum purchases in its chart on U.S. mobile industry “capital investment.” CTIA shows cumulative “capex” between 2016 and 2020 as a cumulative $138.5 billion total.
CTIA also shows spectrum licenses paid for in the 2017 to 2021 period as $116 billion. So obviously, spectrum license spending, though “capex” for accounting purposes, is separated from spending on spectrum licenses, which also is categorized as “capex” by accountants.
Standard and Poors includes spectrum licenses in the “capex” category, for example.
So actual mobile capex often is portrayed as higher or lower, depending on the assumptions. “Network capex” often is separately portrayed from “spectrum capex.” Both have grown.
Among the obvious ways suppliers will hype their own firms is to talk about “6G” networks for which standards have not yet been set. So one hears that “ LG Electronics (LG) successfully demonstrated the transmission and reception of wireless 6G teraHertz (THz) data over 100 meters in an outdoor setting.
That test used frequencies in the 155 GHz to 175 GHz range. Some would say those frequencies are “sub-teraHertz,” not teraHertz, though some consider teraHertz frequencies to range from 100 GHz to 10 THz.
Perhaps most observers consider the millimeter wave spectrum to run between 30 GHz and 300 GHz.
The point is that it is hype that any transmission can be called “6G” at the moment. Apparently we now also disagree about what “teraHertz” means.
Some of us have always seen private 4G and private 5G networks through the prism of local area versus public wide area networking. Which is to say that private networks are always important to different parts of the networking ecosystem.
Private networks are really important for sellers of infrastructure products, less so for suppliers of WAN services. The huge growth of Wi-Fi has huge importance for semiconductor suppliers, systems integration firms and premises networking gear.
Obviously all that use of Wi-Fi also creates needs for WAN networking, but indirectly. The new wrinkle is that 5G private networking creates potential new markets for sellers of public infrastructure and related services such as network design, installation and maintenance.
It has been less clear that private 5G networks represent a big revenue growth opportunity for public network connectivity providers. Some incremental revenue, yes, but nothing in the billions of dollars per service provider range.
So it is that analysts at ABI Research now say “the telco industry needs to radically rethink their approach to enterprise 5G or miss out on the opportunity entirely.”
Perhaps a related issue is that many private 5G networks at present are “demonstration cases” by vendors trying to show the relevance of their solutions. Perhaps not so many are driven “bottoms up” by enterprises with a clear business objective and a perceived existing need.
“In Germany, most private networks are constituted by system integrators or factory automation solution vendors, aiming to showcase 5G capabilities and test solutions to integrate into their product offerings,” says Leo Gergs, ABI Research senior analyst.
“Most private network deployments in Germany are essentially sales-driven and only a few deployments are really used to enhance enterprise workflows and operations,” he notes.
ABI Research argues that the “window of opportunity for enterprise 5G is closing.” The issue is “for whom” is the window closing, if it is. Some of us never believed private 5G networks would “mostly” benefit service providers. Instead, it has seemed--as it the case for other forms of private networking--to be an opportunity primarily for infrastructure providers.
ABI argues that service providers need to reexamine the business model. Others might argue there is not such a compelling argument for a major service provider role. System integrators, for example, have always been important suppliers of enterprise private networks.
There seems nothing so special about 5G private networks that makes public network service providers essential or natural providers.
Maybe service providers should not waste too much effort trying to change that state of affairs.
With the caveat that prices vary from location to location and also when services are bundled, fixed network internet access prices range from about 12 cents per Mbps (or less) on fiber-to-home or cable modem connections to about 25 cents per Mbps on fixed wireless services.
Cost per Mbps generally drops with higher speeds close to 1 Gbps.
Very few observers likely believed that fixed wireless home broadband prices would be that close to fixed network services at intermediate speeds.
Sometimes a bug can be a feature. Perhaps, more commonly, a bug is an implicit end user request for a feature not originally envisioned. In the radio communications world, high signal attenuation is typically considered a problem, hence a “bug.”
Consider atmospheric signal attenuation around 60 GHz. Since high signal attenuation means limited transmission distance, most of us would consider the signal absorption in the atmosphere for signals sent around 60 GHz to be a “bug,” for the same reason signal attenuation at higher frequencies has been viewed as a bug, and why lower frequencies have historically been favored for mobile communications.
source: Signals Research Group, FCC
But high signal attenuation is a bug that becomes a feature for mesh access networks. Basically, the signals attenuate so rapidly that there actually is little signal interference for access networks with many paths.
In other words, the highly-directional signal performance actually provides interference protection.
In perhaps similar fashion, millimeter wave signal attenuation (a “bug”) also comes with a “feature” (capacity 10 times to 100 times greater than signals at lower frequencies). Basically, one trades coverage (distance) for capacity.
For mobile operators facing ever-increasing requirements for capacity, but with stubborn limits on end user willingness to pay, high-capacity millimeter wave spectrum is a feature, not a bug. The same will be true as mobile and wireless communication networks begin to exploit teraHertz spectrum as well.
Fixed wireless always has been a niche platform for access services. But fixed wireless might be significant for suppliers of 5G services as among the few new revenue sources in the consumer market.
Assume there are 75 million 5G fixed wireless connections in service by about 2026. Since most 5G consumer mobility accounts will replace a 4G connection, there is relatively little incremental revenue upside, even if 5G accounts wind up generating higher average revenue than a 4G account, for any reason.
5G fixed wireless connections could--in many cases--provide significant incremental revenue growth, allowing mobile operators to compete for home broadband accounts now largely the province of fixed network operators.
Assume global gross national income gross national income per capita of about $11,600 and a monthly home broadband cost of five percent of GNI per capita. That is about $580 in annual revenue per line.
So 75 million new fixed wireless accounts represents perhaps $43.5 billion in new annual revenue for mobile service providers. That can take the form of new accounts or market share taken from other suppliers.
To put that into perspective, consider projected revenue for other new services. In 2024, it is conceivable that IoT connectivity revenues for mobile operators globally could be in the low millions to tens of millions of dollars, according to Machina Research. Millions, not billions.
In 2026 the global multi-access edge computing market might generate $1.72 billion. Even if one assumes all that revenue is connectivity revenue booked by mobile operators, it still is a far smaller new revenue stream than fixed wireless represents.
In 2020 there were perhaps 80 million fixed wireless subscriptions in service. Researchers at Mobile Experts see that number growing to almost 200 million by 2026.
Ericsson notes that more than 70 percent of all service providers now offer fixed wireless access services. Ericsson also predicts that fixed wireless connections will exceed 180 million by the end of 2026.
By 2026, assuming these forecasts are accurate, fixed wireless will represent about 12 percent of fixed network broadband connections, Ericsson estimates.
Keep in mind that the incremental revenue from 75 million 5G fixed wireless connections does not include the revenue from 4G fixed wireless connections, which might represent another 110 million connections.
That represents an additional $104.3 billion in annual revenue, assuming a global average of $48 per month, per line.
The point is that incremental revenue from 5G fixed wireless is likely to dwarf new revenues earned by mobile operators from edge computing and internet of things.
Some numbers are underwhelming, even when touted as evidence of growth. If you were told that 370 enterprises globally were using local area networks, Wi-Fi, robots, internet-connected sensors, doing local computer processing or were connected to remote cloud computing facilities, would you really be impressed?
Would 370 enterprises globally using computing, software as a service, content delivery networks or some form of artificial intelligence or machine learning capabilities, strike you as terribly significant?
In other words, private 4G or private 5G networks are sort of like that. The use of typically “public network” technology to create a “private network” is important to public network infrastructure providers, as it creates an ancillary new market.
Private 4G or 5G networks might also drive some incremental revenue for public networking companies, who can design, build or operate such networks on behalf of enterprises.
Still, private network revenue is always dominated by infrastructure suppliers and systems integration suppliers. There seems little reason to think private 4G and 5G networks will deviate from that pattern.
Revenues from Wi-Fi as a service likely will still be in single-digit billions globally by about 2025, according to Markets and Markets.
The Wi-Fi infrastructure market, on the other hand, might reach $66.5 billion in 2024, according to Ameri Research. In other words, managed services provided by all suppliers might represent 11 percent of total Wi-Fi sales activity in any given year.
There is a good reason why one of the design goals for 5G was vastly-increased ability to support huge device density. 5G was intended to support internet of things sensors that often would be deployed in huge numbers.
Consider present use of IoT devices in hospitals. A survey by Philips found that device populations can easily exceed 1000 to 10,000 devices at a single location. Some 56 percent of locations already have more than 10,000 devices in operation, up to 50,000 or more.
4G networks can handle a maximum of about 100,000 devices per square kilometer (an area of about 0.46 miles). 5G is designed to handle a million devices per square kilometer.
Lots of people who know a bit about radio frequency networks will justifiably have some concern about how well commercial mobile systems will function when using GHz frequencies between 30 GHz and 140 GHz, for example.
All radio frequency signals suffer high attenuation in the first few meters, once they are transmitted. This effect is essentially frequency independent.
In other words, radio frequency signals get weaker according to an inverse square law: quantity is inversely proportional to the square of the distance from the source: 1/distance squared.
Attenuation in free space also is increased by rain or snow, and is frequency dependent.
So as we start to commercially deploy radio systems between 10 GHz and 400 GHz, we will see higher free space attenuation than at sub-1 GHz frequencies or mid-band frequencies (mid-band roughly between 2 GHz and 6 GHz).
Higher frequencies with higher free space attenuation is why small cells are important for millimeter wave signals (26-, 28-, 38-, and 60 GHz in the U.S. market).
But some are optimistic that as we begin to deploy mobile generations beyond 5G, and open up millimeter and teraHertz frequencies for use, we will find affordable ways to deploy radio infrastructure that is commercially useful.
One big concern is signal propagation. It might ultimately be less a problem than we imagine.
“Between sub-6 GHz and 140 GHz frequencies, the propagation path loss for an urban radio channel doesn't really differ at different frequencies, after accounting for the radiated signal's first meter of travel,” says Theodore S. Rappaport, Professor of Electrical Engineering at the NYU Tandon School of Engineering.
This means that once a radio signal reaches what's called the "far field" (beyond the first meter or so), the frequency has surprising little impact on a signal's attenuation as it travels through urban and indoor channels.
The caveat is that rain will increase attenuation. And signals at some frequencies are more liable to be absorbed by oxygen. That is a big deal just beyond 60-GHz and around 120 GHz. In those two regions point-to-point operations are possible, as well as indoor, short range use cases.
Rappaport believes that future mobile networks using frequencies above 100 GHz will likely work, at most of the same locations. “Networks won't likely require further densification, and today's new tower sites for 5G will be usable for decades to come without the need to build many more,” he argues.
Beginning with 5G, wireless systems are using directional antennas that have high antenna gains and narrow beam widths on both the mobile and the base station ends of each link, Rappaport says. This offers more, not less, signal strength to each user as we move to millimeter wave, sub-teraHertz, and ultimately teraHertz frequencies.
Good news, if he proves right for networks built at scale.
