Wide Divergence on 5G, IoT, Edge Computing Near Term Impact

It is not too hard to find skeptics about revenue upside for mobile operators in 5G, internet of things or edge computing, because it is hard to demonstrate the business case in a near term sense, in most markets. That seems to have been true at the recent Mobile World Congress trade show as well.

Despite clear evidence that mobile operators in the United States, Japan, South Korea and China are charging fast into 5G, with firms in each of those countries believing 5G leadership will lead to broader benefits, many operators in other nations are less optimistic about the near term business case.

In some cases, that skepticism or agnosticism also has other roots: recent and major investments in 4G, uncertain spectrum allocations or the general state of the mobile business (very low growth) in general, which raises the risk of big new investments in 5G.

Add to that divergence the sense in early-adopter countries that 5G is about industrial policy, while perhaps not a plausible possibility in most countries, and the split of opinion is easy to explain.

Optimists might point out that business model fuzziness also was a characteristic of both 3G and 4G, so some amount of uncertainty is perhaps typical of next-generation mobile platforms in recent eras that have seen revenue growth drivers migrate from voice to messaging to internet access for revenue growth, and which now will have to depend on additional new growth drivers in the 5G era beyond internet access for people on smartphones.

It also is fair to say that early adopters in 5G vary by continent, with mobile operators in the United States, Japan, China, South Korea moving early and aggressively into 5G, while European, African, South American and broader Asian operators are more circumspect.

In some instances, the hesitation is related to the recency of 4G investments. In other cases the difficulty of the mobile business model in general makes a major upgrade cycle unpalatable.

There are divides, by continent, on the matter of internet of things upside as well, possibly because many observers see the biggest early revenue cases developing in North America and parts of Asia.

source: Research Nester

So it might not be surprising that mobile operators in the United States and East Asia are more optimistic about the near term 5G and IoT business cases (as IoT upside will be dependent on 5G, in important cases).

In Europe, the IoT opportunity often is included in discussions of the “fourth industrial revolution.”

One hears the term of art used less in the North American markets, perhaps suggesting different perceptions of where opportunity lies. In a broad sense, users of the term fourth industrial revolution seem to use it in a sense similar to “digitalization” and a bit more reliant on applying artificial intelligence to business processes.

Other touted revenue platforms such as edge computing are simply early in development, as edge computing requires new lead apps such as autonomous vehicles to drive such requirements.

Some observers also note the nebulous discussions and status of artificial intelligence within the service provider community at the moment, though, perhaps surprisingly, AI is seen by some as having more relevance for mobile operators than 5G, IoT or edge computing.

The point is that near-term optimism about 5G, internet of things and edge computing--if not complete conviction at the same level--is higher in a few markets than across most of the globe.

In part, that is because 5G business cases are part of the broader IoT and edge computing trends. Without significant IoT growth, it is hard to make the case either for 5G or edge computing.

At the moment, 5G and IoT are a Rorschach test. Different people see different things.

Facts Do Not Support Charge that Internet Access Prices Climb "Without Net Neutrality"

One often hears, in discussions of network neutrality, assertions that without such rules, internet access costs will skyrocket. Oddly enough, internet access has--for nearly all its history--been an unregulated service where internet service providers have been able to charge whatever the market will bear.

Since, for all of the time U.S. consumers have purchased internet access, the same fundamental rules (internet access is a data service, and therefore unregulated as are other data applications) have been in place, with the exception of the the few years when internet access was deemed to be a regulated “telecom” service, we can look at the whole history of prices to evaluate whether prices actually have shown evidence of “price gouging.”

History suggests prices have done nothing but drop--and drop by orders of magnitude--over the whole history of consumer internet access services.

Some might argue that this might be true on a “cost per bit” basis. Others will argue it is true on an absolute price basis. Those of you who were buying internet access early in the broadband era (around the turn of the century) can remember spending $200 to $300 a month for 756 kbps service. We routinely now pay $50 to $70 a month for service of 100 Mbps.

Value is higher and price is lower, on an absolute basis, without adjusting for inflation. And prices continue to fall, even faster, on a “value compared to price” basis. Schools buying internet access between 2013 and 2017, for example, saw prices decline 78 percent, on a cost per Mbps basis.

source: Infrastructure USA

That same price trend can be seen globally, in developing countries. The same trend was seen in mobile voice, internet transit (where prices often decline 33 percent per year).

Consumer internet access also can be characterized as having Moore’s Law rates of change. In other words, connection speed increases 50 percent per year, every year, while prices drop, remain stable or increase only modestly.

source: Statista

It is obvious that consumer internet access prices, despite improvements, do not increase 50 percent per year. In fact, without adjusting for inflation or changes in end user demand for products, it is hard to argue that U.S. consumer internet access prices actually have increased much at all.

Real prices are often hidden because most consumers do not buy the “stand alone” prices for access, but instead buy bundles of services where the actual discounts are not visible.

Some might argue internet access prices rise two percent or perhaps three percent a year. That is not surprising, as this represents the underlying rate of inflation.  

source: Jacob Nielsen    

The point is that actual access prices are not--as many claim--increasing in the U.S. market because there is no competition or because there is no “network neutrality.”

Some studies suggest prices are higher than they actually are, because the quotes are for “standalone” prices, at retail, even when most consumers do not buy those products, but rather bundles or other promotional plans.

Of course, one also has to adjust nominal (actual) retail prices across countries, as overall price levels vary across countries. Adjusting using the purchasing power parity approach, to normalize retail prices for actual price levels in each country, global internet access prices average between $50 to $80 a month, according to Point Topic studies.

source: Point Topic

Will SD-WAN Become the Universal Enterprise WAN Connection?

Some 93 percent of surveyed executives believe enterprises will have adopted software-defined wide area networks by the end of 2019.

That is a typical, and possibly too-optimistic forecast, in terms of revenue expectations. Since SD-WAN primarily is viewed as a branch office solution, and since current revenue might be less than $1 billion annually, on a global level. That is a very-small market, by tier-one service provider standards.

But there is another way to look at what we now call “software defined wide area networks.” And that is to envision the future enterprise wide area network being built nearly exclusively on virtual private networks that tunnel through the public network.

Source: Nokia Bell Labs

source: Nokia Bell Labs

In that scenario, virtual private networks become the basic way all enterprise WAN connections are built--to headquarters or branch locations. That becomes even more central if, as many believe, future networks will be highly virtualized and therefore federated.

The Whole Next Generation Network Will Feature Ultra-Low Latency

Many observers probably agree that 5G really is about networks that have ultra-low latency and gigabit access speeds. Most would agree 5G is a mobile network implementation of a next-generation network.

What fewer might recognize is that the whole network (WAN, metro and access) is changing. The whole network fabric (across WAN, metro and access) will be based on ultra-low latency and gigabit bandwidth.

The reason is that the next huge wave of revenue growth, apps and services is expected to be based on applications that require ultra-low latency, and sometimes gigabit speeds, though that is the lesser of the requirements for most apps.

Network architectures always directly reflect perceptions of what drives revenue, and the huge bet to be made is that low-latency apps will drive incremental revenue growth, not the traditional mobile or web services consumed by humans on smartphones.

The corollary is that edge computing centers will emerge as a key new part of the networking architecture.

Many decades ago, the architecture of a telecom network, in a single domain, was simple.  

Traffic passed between active switch nodes, within and across across domain boundaries.

The business model was simple: people making phone calls, so the network was optimized for phone calls.

source: Web Classes

These days, a more-relevant diagram would focus on the packet network, since that now is the way most people interact with “telecom” networks. The core network largely remains the same, with traffic passed between provider domains at a packet network gateways or switches.

The legacy fixed network remains, but most of the revenue now flows over the IP networks (mobility, enterprise communications, video and other internet apps).

Along with new architecture, revenue models have changed. Voice still drives revenue in developing markets, but in developed markets voice and messaging are mature or declining revenue sources. Growth now comes from internet access and video entertainment.

Devices now include phones, but also feature sensors, PCs and tablets.

The more-important change is the separation of control (signaling) and bearer traffic (content, voice, messaging); the separation of transport/access and content sources and the “openness” of the whole network to third party access/traffic.

Technologists call this a separation of control plane (signaling and control) and data plane (bearer traffic, content).

Where once the telecom network was closed, it now is open. Where once the telecom service provider tightly controlled permissible apps and devices, now admission to the network is open to all third parties who comply with the network protocols and provide lawful applications.

source: GetGo Technology

What comes next might be different as well. As internet protocol has become the universal next generation protocol, all networks have become computing networks.

And most computing networks now are cloud based, for consumer apps, and for a majority of enterprise apps.

source: Cisco

That reliance on cloud computing is predicted to grow in the coming 5G and internet of things era, primarily because many new apps require ultra-low latency.

But something new is coming. Some important cloud-based apps and revenue drivers will require ultra-low latency, meaning centralized cloud computing centers will not work. Instead, computing will have to be done at the edge.

Source: Nokia Bell Labs

That architecture will be built to support new applications and revenue drivers dependent on ultra-low latency (more than bandwidth, even if bandwidth will be in gigabit ranges).

In the Next Eras of Communications, All Access will be Wireless (Untethered)

Network architectures and communication protocols are not the primary concern of most people who work in and around the telecommunications industry, but architectures always reflect the dominant business models in each era.

One way of stating the principle is that networks always have been optimized for the lead application and revenue driver, in each era. Broadcast TV and radio; cable TV and satellite video networks as well as telecom networks were cases of “form follows function:” each network was designed to optimally support one key app.

We can argue that all began to change when next-generation networks were envisioned as “all digital,” allowing one platform to support any media type. That is generally--if not entirely--true.

Cabled networks are designed to connect places or locations; mobile networks are designed to connect people or devices.

But we are likely headed for another evolution of network architectures, precisely to support new potential business models. Broadly, all networks these days are computing networks.

So if computing application requirements change (along with revenue drivers), then architectures will evolve to supply that demand. Mainframe, minicomputer, personal computing, cloud computing and mobile computing all changed networks.

If the next evolution of demand for computing--and networks to support computing--centers on sensors, not people, and if important categories of such computing must feature ultra-low latency, then centralized cloud computing will not work.

A new network and computing architecture will have to be created “at the edge” to support local processing. As always, “what computing” has to be supported changes “where” computing happens.

source: Nokia Bell Labs

High-frequency trading, virtual reality gaming, autonomous vehicle, augmented reality gaming,  4K video and remote medical apps are among conceivable apps that require such low latency that centralized cloud computing will not work.

Ultra-low latency, even with ultra-low-latency access networks, might require edge computing as well, as the time to reach remote computing resources is too lengthy.

That “ultra-low latency” and “gigabit” access speeds are primary characteristics of coming 5G networks can obscure the larger implications. All networking is moving to “ultra-low latency” and gigabit speeds.

The importance of 5G is not just that it is the next generation of mobile communications. The larger point is that 5G is part of the re-architecting of networks in general towards ultra-low latency, high performance support where the core network resources are accessed using untethered means.

That will be true for consumer or enterprise access. The whole point of access networks will be to support ubiquitous access to core network resources over “long wires and short wireless connections.”

In that sense, “5G” is shorthand. It represents not only a particular mobile solution for untethered access, but a generic shorthand for all ultra-low latency untethered access networks as well.

That is imprecise, of course. But one stumbles over longer phrases such as next-generation networks, which in turn require explanation. The term “5G” (although a specific mobile solution) necessarily highlights the coming change in network architectures and business models.

For the first time, some important revenue-generating applications will require such low latency that computing resources (cloud data centers) must be moved to the edge of the network.

Traditional web apps, voice and messaging still will be able to use centralized cloud data centers. But many of the hoped-for new apps will have to move to the edge of networks. And no matter what the setting (consumer or business; indoors or outdoors; home, small business or large campus), access will be on an untethered basis.

Mobile, Wi-Fi, or faster protocols such as Wi-Gig will be used. But all will be untethered (no wires).

Many Believe Artificial Intelligence Will Destroy More Jobs than it Creates

Artificial intelligence cannot fail to reshape work. Many workers fear (rightly or wrongly) that AI will replace them. More optimistic observers rightly argue that new jobs will be created.

But we have seen big changes in economic eras before. And one clear danger is that the people losing jobs to AI will not be the people getting the new AI-related jobs. In the transition from an industrial to an information economy, big and sustained job losses have occurred in some regions and industries, to some specific workers.

There is plenty of evidence that many of those workers never again were able to replace the lost income or jobs. Among the biggest losers in the U.S. economy are males 25 to 54. Labor force participation for people in that demographic has been dropped steadily since 1950.

That 70-year trend might well add new demographics as we make the transition from an early-stage information economy to a middle-age information economy where artificial intelligence is applied to most activities, as was the case for computing and software.

About 74 percent of polled workers in India agree that AI is better than people for at least some tasks. About 58 percent of surveyed U.S. workers agree that AI is better than people at some tasks.

source: eMarketer

Fully 73 percent of U.S. respondents to a Gallup poll believe AI will reduce more jobs than it creates.

source: eMarketer

Telcos Look to AI to Reduce Costs

One thing about ecosystems and value chains is important: one segment's cost is another segment's revenue. That happens at the firm level as well.

Artificial intelligence, many fear, will eliminate and reshape many, if not most, jobs. Not to worry, optimists say, new jobs will be created. While that is true, the social consequences might be anything but benign.

Telecom service providers, for example, now expect that artificial intelligence will allow companies to reduce headcount, boosting productivity but at the cost of jobs. That always is the challenge with higher productivity. The ability to produce more for less drives economic growth, but has uneven implications for workers.

Consider the changes in job markets over the last six or seven decades, as the United States moved from an industrial to information-driven economy.

The labor force participation rate of prime-age men--25 to 54 years old--steadily decreased since 1950.

To be sure, major economic shifts do eliminate some jobs and create others. The social problems happen because the people who lose old jobs are quite often not the people gaining the new jobs.

source: Bureau of Labor Statistics

The point is that smaller workforces have been a service provider objective for decades. AI is going to push the trend.