Definitions Matter

Minimums, median and maximum all are valuable indices in life, business and nature, including measures of internet access adoption or “quality.” It also has to be noted that constantly moving goalposts--changing our definitions--is a way of creating permanent problems.

That is not to deny the usefulness of revising our definitions over time. It is a truism that yesterday's power user is today's typical user. 

The percentage of U.S. customers buying internet access at the minimum speeds keeps dropping, as customers migrate to tiers of service that offer higher speeds at the same or only slightly-higher cost. 

But such definitions matter for both consumers and suppliers. Customers might sometimes buy services that actually are overkill. Most internet access customers buy what they believe is good enough to support their actual use cases, and rarely what is the “best” available level of service. 

Suppliers might imperil their business models by forcing investment in facilities that customers will not use, overprovisioning service in ways that raise sunk costs of doing business without providing capabilities customers actually buy. 

Those buying patterns also suggest why some ISP offers that are not state of the art can still be commercially viable. The reason is that, beyond a certain point, additional speed provides no tangible user experience benefit.  

And permanent problems are essential for those who claim to be in the business of “solving those problems.” That matters for education, health, disease, economies, social and economic equality, sports and just about anything else you can think of. In other words, one cannot marshall public policy support to solve a problem that does not exist.  

To be sure, our definitions of “broadband” have evolved, and will continue to evolve. 50 years ago, broadband was defined as any speed at 1.5 Mbps or faster. Once upon a time, Ethernet ran at 10 Mbps, while fiber to the home offered 10 Mbps. Today’s systems all run much faster than that. 

But it also makes a difference to “problem solvers” that definitions are revised upwards. Doing so always creates a “bigger problem.” 

Changing the minimum definition of broadband would shift the size of the “underserved” population or locations, for example. Today, perhaps 20 percent of U.S. buyers of fixed network internet access purchase services at the minimum speed of 25 Mbps. 

Changing the definition to 100 Mbps would increase the size of the underserved locations to nearly half of all buyers. Again, we will keep increasing both minimum levels of service, customers will keep changing the speed tiers they purchase and internet service providers will keep supplying faster speeds. 

The point, however, is that changing minimum definitions does not change the number or percentage of tiers of service customers purchase or that ISPs supply. Already, we find that the percentage of customers buying the fastest-possible speeds (at least 1 Gbps) is in mid single digits. 

More to the point, the typical buyer prefers a service offering 100 Mbps to 400 Mbps. Changing “minimum” to “average” has consequences, arguably distorting our understanding of “good enough” levels of broadband speed. 

Benchmarks are valuable when trying to measure “progress” toward some stated goal. A minimum speed definition for broadband access is an example. But that does not obviate the value of knowing maximum and median values, either, especially when the typical U.S. internet access buyer routinely buys services significantly higher than the minimum. 

In the first quarter of 2020, for example, only about 18 percent of U.S. consumers actually bought services running at 40 Mbps or less. All the rest bought services running faster than 50 Mbps. 

source: Openvault

An analysis by the Open Technology Institute concludes that “consumers in the United States pay more on average for monthly internet service than consumers abroad—especially for higher speed tiers.” 

As always, methodology matters. The OTI study examines standalone internet access plans, even if that does not account for the plans most consumers actually buy. The figures do not appear to be adjusted for purchasing power differences between countries. Were that done, it might be clearer that average internet access prices are about $50 a month, globally. 

Global prices are remarkably consistent, in fact, when adjusting for purchasing power conditions in each country.  

Nor does any snapshot show longer term trends, such as lower internet access prices globally since at least 2008. A look at U.S. prices shows a “lower price” trend since the last century. U.S. internet access prices have fallen since 1997, for example. 

source: New America Foundation

The OTI study claims that, comparing average prices between markets with and without a municipal system shows higher prices in markets with government-run networks. Not all agree with that conclusion. 

“The OTI Report’s data, once corrected for errors, do not support the hypothesis that government-run networks charge lower prices,” says Dr. George Ford, Phoenix Center for Advanced Legal and Economic Public Policy Studies chief economist. 

“Using OTI’s data, I find that average prices are about 13 percent higher in cities with a municipal provider than in cities without a government-run network,” says Ford. 

Our definitions of “broadband” keep changing in a higher direction. Once upon a time broadband was anything faster than 1.5 Mbps. Ethernet once topped out at 10 Mbps. 

Today’s minimum definition of 25 Mbps will change as well. The point is that having a minimum says nothing about typical or maximum performance.

About 91 percent to 92 percent of U.S. residents already have access to fixed network internet access at speeds of at least 100 Mbps, according to Broadband Now. And most buy speeds in that range. 

source: Broadband Now

It is useful to have minimum goals. It also is important to recognize when actual consumers buy products that are much more advanced than set minimums. 

Regulators Cannot Keep Up with Pace of Change in Computing and Communications

It often has been said that regulators cannot keep up with the pace of change in computing, broadband and applications. That has proven to be true for regulators looking at home broadband performance. 

About 2010 or so Ofcom, the U.K. regulator laid out a national goal for “superfast” internet access of about 30 Mbps, at a time when the typical speed most consumers were able to use was about 6 Mbps.

Average actual U.K. fixed-line residential broadband speeds grew from about 3,6 Mbps in 2008 to about 15 Mbps in 2013. 

source: Ofcom 

In 2011, Ofcom warned of low interest in 50-Mbps services, for example. 

Ofcom also worried about low interest in 30 Mbps services as well. 

That same year, a 50-Mbps internet access connection (home broadband) cost close to $100 a month. 

The next formal goal will be gigabit per second access, which shows you just how fast improvements are coming in the home broadband business. 

That same degree of improvement was seen in the U.S. market as well. Back in 2010 average U.S. home broadband speeds were about 5 Mbps. By 2019 speeds had climbed to about 33 Mbps. 

U.S. median home broadband speeds were about 131 Mbps in October 2021.  

source: Nielsen Norman Group 

By 2050 the home broadband headline speeds are likely to be in terabits per second. Though the average or typical consumer does not buy the “fastest possible” tier of service, the steady growth of headline tier speed since the time of dial-up access is quite linear. 

And the growth trend--50 percent per year speed increases--known as Nielsen’s Law--has operated since the days of dial-up internet access. Even if the “typical” consumer buys speeds an order of magnitude less than the headline speed, that still suggests the typical consumer--at a time when the fastest-possible speed is 100 Gbps to 1,000 Gbps--still will be buying service operating at speeds not less than 1 Gbps to 10 Gbps. 

Though typical internet access speeds in Europe and other regions at the moment are not yet routinely in the 300-Mbps range, gigabit per second speeds eventually will be the norm, globally, as crazy as that might seem, by perhaps 2050. 

The reason is simply that the historical growth of retail internet bandwidth suggests that will happen. Over any decade period, internet speeds have grown 57 times. Since 2050 is three decades off, headline speeds of tens to hundreds of terabits per second are easy to predict. 

source: FuturistSpeaker 

Some will argue that Nielsen’s Law cannot continue indefinitely, as most would agree Moore’s Law cannot continue unchanged, either. Even with some significant tapering of the rate of progress, the point is that headline speeds in the hundreds of gigabits per second still are feasible by 2050. And if the typical buyer still prefers services an order of magnitude less fast, that still indicates typical speeds of 10 Gbps 30 Gbps or so. 

Speeds of a gigabit per second might be the “economy” tier as early as 2030, when headline speed might be 100 Gbps and the typical consumer buys a 10-Gbps service. 

source: Nielsen Norman Group 

Europe, U.S. Markets On Track for Widespread 100 Mbps by 2024

Since 2010, regulators in the United States and European Union have called for dramatically-faster Internet access, both targeting speeds of 100 Mbps by about 2020, in the case of the EC.

The U.S. Federal Communications Commission’s National Broadband Plan calls for providing at least 100 million U.S. homes with “affordable access” to actual download speeds of at least 100 megabits per second and actual upload speeds of at least 50 megabits per second, without setting a specific time frame.

At the time both initiatives were launched, the goals might have seemed farfetched. In 2010, according to Akamai, typical U.S. and European access speeds were about 4 Mbps, though some studies showed higher speeds.

That we now have major Internet service providers talking about, and in some cases, building networks capable of supplying gigabit Internet access shows how fast supplier thinking has changed since 2010, and how fast higher speeds are being made available, despite a persistent sense in some quarters that progress is way too slow.

In 2002, most U.S. households did not even have access at 1.5 Mbps. By 2013, according to Akamai, typical U.S. speeds were about 10 Mbps, showing roughly an order of magnitude increase over a decade. By that metric, 100 Mbps should be what a typical user buys by about 2024.

The European Commission’s Connected Continent initiative reports that broadband availability now is 100 percent across the EC region, with consumers having multiple choices of service providers.

Access speed, though, remains an issue. People able to use 4G mobile Internet access also rose to 59 percent, up from 26 percent a year ago.

Fixed network Internet access operating at 30 Mbps or higher is available to 62 percent of the EU population, up from 54 percent  a year ago and just 29 percent in 2010.

Fast broadband is already available to 90 percent of homes or more in Belgium, Denmark, Lithuania, Luxembourg, Malta, the Netherlands and the United Kingdom.

As you would guess, the biggest problems are rural areas, where 18 percent of rural households have access high-speed broadband access.

The current objective is downstream speeds of 30 Mbps for everyone in the EC region and at least 50 percent of European households subscribing to Internet connections above 100 Mbps by 2020.

Those goals were announced in 2010.

Standard fixed broadband now covers 95.5 percent of EC homes, while rural coverage of standard fixed broadband was 83 percent at the end of 2012, to an EC broadband report reporting status up to July 2013.

Connections capable of providing at least 30 Mbps download cover 54 percent of EU homes. Cable has the highest coverage, at 39 percent of homes, followed by  very high speed digital subscriber line at 25 percent and fiber to the home at 12 percent of homes.

What in the United Kingdom is called “superfast” (30 Mbps or faster) access accounts for 20 percent of all fixed broadband lines in the EC, as opposed to 12 percent a year ago. Some 57 percent  of such connections are supplied by cable TV operators.

In fact, new entrants provide 77.5 percent of faster connections.

About two percent of homes buy fixed network service operating at 100 Mbps or faster.

About 15 percent of homes buy service at 30 Mbps or faster.

Though consumer behavior might have suggested there was little demand for 50 Mbps or 100 Mbps Internet access in the past, the primary reason for limited demand was the cost. As gigabit access network pricing has been redefined, to about $70 or $80 a month in the U.S. market, for example, demand is at least as high as for today’s more common 20 Mbps to 40 Mbps services.

Digital Redlining or Response to Demand?

Terms such as digital redlining imply that U.S. internet service providers upgrade neighborhoods able to pay for higher speed internet access underinvesting in poorer neighborhoods. At some level, it is hard to argue with that point of view, at least where it comes to gigabit internet access. 

Google itself pioneered the tactic of building in neighborhoods where there is demonstrated demand, building Google Fiber first in neighborhoods (typically higher-income areas) where potential customers were most interested. Other gigabit service providers have used the placing of deposits for the same reason. 

And regulatory officials at the local level seem to now agree that “universal service” (building a gigabit network past every home and business) is desirable in some cases, but not absolutely mandatory in all cases. The thinking is that allowing new internet service providers or facilities to be built wherever possible is a better outcome than requiring ubiquity, and getting nothing. 

Also, higher-speed facilities often are not found everywhere in a single market or city. CenturyLink does sell gigabit internet access in Denver, just not everywhere in the metro area. That is not necessarily “redlining,” but likely based on capital available to invest; expectations about financial return; customer density or any other combination of business issues that discourages the investment in new access facilities. 

The economics of communication networks also are clear. Density and cost per location are inversely related. Mobile networks typically have 10 percent of cell sites supporting 50 percent of usage. About 30 percent of sites carry about 80 percent of traffic. That has been true since at least the 3G era.  

In fixed networks, network cost and density also are inversely related. So population density has a direct bearing on network costs. In the U.S. market, network unavailability is concentrated on the last couple of percent of locations.  

With cable operators already holding at least 70 percent share of the internet access installed base of customers, any new investment in faster facilities faces a tough challenge. Any new fiber to home network, for example, essentially is playing catch-up to a cable operator, as roughly 80 percent of U.S. households already also are reached by gigabit speed cable networks. 

And cable share has grown, up from possibly 67 percent share in 2017. 

That noted, internet speeds do vary by geography: speeds in urban areas frequently are higher than in rural areas. But the argument that large numbers of U.S. households are underserved often is correct, depending on what standard one wished to apply, and how one defines the supplier market.

Some claim 42 million U.S. residents are unable to buy broadband internet access, defined as minimum speeds of 25 Mbps in the downstream.  That actually is incorrect. 

Virtually every household in the continental United States is able to buy 25 Mbps or faster service from at least two different satellite providers. But those who claim “42 million” people cannot buy broadband simply ignore those choices, and focus only on the claimed availability of 25 Mbps service by fixed network providers. 

There are other estimates which also vary wildly. Roughly 10 percent of U.S. households are in rural areas, the places where it is most expensive to install fast fixed network internet access facilities, and where the greatest speed gaps--compared to urban areas--almost certainly continue to exist.

In its own work with TV white spaces, Microsoft has targeted perhaps two million people, or roughly a million households, that have no fixed network internet access. That assumes there are two people living in a typical household, which is below the U.S. average of roughly 2.3 to 2.5 per household.

Recall that the definition of broadband is 25 Mbps downstream. Microsoft has argued that 20 million people (about 10 million homes) or perhaps eight percent of the population (perhaps four percent of homes) cannot get such speeds from any fixed network service provider.

Microsoft also has cited figures suggesting 25 million people cannot buy broadband--presumably using the 25 Mbps minimum standard, most of those people living in rural areas. 

That conflicts with data from Openvault that suggests 95 percent of the U.S. population can buy internet access at a minimum of 25 Mbps, while 91 percent to 92 percent can buy service at a minimum of 100 Mbps. 

Using the average 2.5 persons per U.S. household average, that suggests a universe of about 10 million U.S. homes unable to purchase internet access at 25 Mbps from a fixed network supplier, in 2018. What is not so clear is the percentage of households or persons who can do so using a mobile network. 

None of that explains urban areas with slow speeds, though. There the issue is more likely to be high construction costs in urban areas where underground construction is necessary, along with demand expectations that are lower than in suburban areas. That is true whether it is electrical lines or communications networks being considered.   

But at least one Microsoft analysis suggests that about half of all U.S. households are not using 25 Mbps access. The claim is that 162.8 million people are “not using the internet at broadband speeds.” That seems to clearly contradict data gathered by firms such as Ookla and Opensignal suggesting that average U.S. speeds are in triple digits.

In 2018, the average U.S. broadband speed was 94 Mbps, according to the NCTA. That same year, Ookla reported the average U.S. speed was 96 Mbps. 

It is not quite clear how the Microsoft data was generated, though one blog post suggested it was based on an analysis of “anonymized data that we collect as part of our ongoing work to improve the performance and security of our software and services.” 

The claim of 162.8 million people “not using the internet at broadband speeds” (probably using 25 Mbps as the definition) equates to about 65 million households, using the 2.5 persons per household definition. That does not seem to match other data, including the statistics Microsoft itself cites. 

What remains difficult, but might explain the divergence, is if applications and services include both apps run on smartphones as well as PCs and other devices connected to fixed networks. That would explain the number of users, while usage on mobile networks might account for large numbers of sessions where 25 Mbps speeds downstream were not noted, or perhaps it was the upstream speed definition (minimum of 3 Mbps) that was the issue.  

Even then, downstream average 4G speeds in 2018 were in excess of 40 Mbps downstream, so even that explanation is a bit difficult. 

Perhaps there are other ways to make sense of the data. There is a difference between users (people) and households. There is a difference between usage and availability; usage by device (mobile, PC, tablet, gaming device, sensor); application bandwidth and network bandwidth. 

Perhaps the issue is application performance on a wide range of devices including mobiles and untethered devices using Wi-Fi, which would reduce average experienced speeds, compared to “delivered access speed.” 

Methodology does matter. So do the costs and benefits of broadband capital investment under competitive conditions, in areas with high construction costs or low demand for advanced services, especially when newer platforms with better economics are being commercialized. 

Telecommunications is a business like any other. Investments are made in expectation of profits. Where a sustainable business case does not exist, subsidies for high-cost areas or universal service support exist. 

The point is that every human activity has a business and revenue model: it can be product sales, advertising, memberships, subscriptions, tax support, fees, donations or inheritances. Children have a “parents support me” revenue model, supported in turn by any of the aforementioned revenue models. 

But every sustainable activity has a revenue model, direct or indirect. The whole global communications business now operates on very different principles than the pre-competitive monopoly business prior to the 1980s. We still have a “universal service” low end, but we increasingly rely on end user demand to drive the high end. 

Our notions of low end change--and higher--over time. We once defined “broadband” as any data rate of 1.544 Mbps or higher. These days we might use functional definitions of 25 Mbps or 30 Mbps. Recall that 30 Mbps--in 2020--was called “superfast” as a goal for U.K. fixed network broadband. 

Few of us would consider 30 Mbps “superfast” any longer. Some might say the new “superfast” is gigabit per second speeds. But that is the change in real-world communications over just a decade. What was a goal in 2010 now is far surpassed. 

What some call “redlining” is simply a response to huge changes in the internet access and communications business. “Maximum” is a moving target that responds to customer demand. “Minimums” tend to be set by government regulators in search of universal service. 

As independent internet service providers cherry pick service areas where they believe the greatest demand for gigabit per second internet access exists, so do incumbents. 

Similar choices are made by providers of metro business services; builders of subsea connectivity networks or suppliers of low earth orbit satellite constellations and fixed wireless networks. They build first--or pick customer segments--where they think the demand is greatest.

Faster Home Broadband Just Keeps Coming, As Edholm and Nielsen Laws Predict

Google Fiber will launch 5-Gbps and 8-Gbps internet access service in early 2023. Both products will offer symmetrical upload and download speeds, Google says. 

Google Fiber launched gigabit service in 2010, 2-Gbps service in 2020 and (and is testing 20-Gbps service. 

The 5-Gbps tier will cost $125 per month, while the 8-Gbps tier will cost $150 per month. 

Separately, it seems increasingly likely that Comcast will begin to introduce service at speeds possibly in the 4-Gbps to 6-Gbps range in 2023. And those services might well be symmetrical, able to extend to 10-Gbps symmetrical. 

Those speed increases are predictable and expected according to two theorems. 

Nielsen's Law suggests that the top-end speed will grow 50 percent per year. Edholm’s Law states that internet access bandwidth at the top end increases at about the same rate as Moore’s Law, which is about a doubling every 18 months or so. 

That means the top-end home broadband speed could be 85 Gbps to 100 Gbps by about 2030. 

source: NCTA  

Nielsen Norman Group estimates suggest a headline speed of 10 Gbps will be commercially available by about 2025, so the commercial offering of 2-Gbps and 5-Gbps is right on the path to 10 Gbps. 

AT&T, for example, just activated symmetrical 2-Gbps and symmetrical 5-Gbps service for 5.2  million locations across 70 U.S. markets, with plans to deploy across the whole footprint in 2022 and later years. 

There is widespread expectation that the headline speed for home broadband, in many markets, will be 10 Gbps by about 2025. 

By other rules of thumb, that also suggests the "typical" home broadband customer will be buying service at rates between 1 Gbps and 2 Gbps, with a significant percentage buying service at 4 Gbps. 

Nielsen’s Law has operated since the days of dial-up internet access. Even if the “typical” consumer buys speeds an order of magnitude less than the headline speed, that still suggests the typical consumer--at a time when the fastest-possible speed is 100 Gbps to 1,000 Gbps--still will be buying service operating at speeds not less than 1 Gbps to 10 Gbps.  

source: FuturistSpeaker 

So top end speeds in the terabits per second are virtually inevitable by about 2050. The emergence of offers between 2 Gbps and 5 Gbps now is simply evidence that the trend continues. 

At the moment, top speeds in the U.S. market are in the 2 Gbps and 5 Gbps ranges.Comcast has introduced 3-Gbps services for business. Ziply has introduced symmetrical 2-Gbps service. Google Fiber has added 2-Gbps as well. 

Frontier Communications is doing the same. Verizon offers 2-Gbps Fios service. AT&T sells both 5-Gbps and 2-Gbps service. Many of those offers feature symmetrical bandwidth. 

Perhaps the greatest value change, though, is not the headline downstream speed, but the symmetrical speeds, as in the U.S. market asymmetrical services sold by cable operators have nearly 70 percent market share. 

Though the cable hybrid-fiber coax networks can be configured to support more upstream bandwidth, fully-symmetrical service typically requires switching to fiber-to-the-home platforms. 

To scale new capital investments, cable operators in many cases will choose to extend downstream speeds and lift upstream speeds, approaching or reaching fully symmetrical service with DOCSIS 4.0 before considering other measures such as switching to FTTH. 

If the “typical” customer buys a service operating at up to an order of magnitude less than the highest headline speed, we might infer that the typical home account--offered by ISPs with various speed plans--will be buying service at speeds between 500 Mbps and 800 Mbps in 2025. 

Keep in mind that Google Fiber’s footprint is quite limited, so not many households will be able to buy Google Fiber service, now generally available at either 1-Gbps or 2-Gbps speeds. In such cases, the headline speed and the median speed tend to be virtually identical. 

The real local market test will tend to be the 2-Gbps to 5-Gbps services sold by Comcast, which has the biggest home footprint, or AT&T, with perhaps the third-biggest footprint. But those services are marketed mostly to business customers, at this point.