Wednesday, March 14, 2018

AT&T Time Warner Acquisition Could Trigger Merger Wave, or Quash It

Consolidation tends to come in waves, as big deals by a major competitor trigger competitive responses by other contestants. So approval--or rejection--of the AT&T acquisition of Time Warner will matter.




Tuesday, March 13, 2018

Core Network Virtualization and the "Dumb Pipe" Reality

Virtualization and “dumb pipe” are related concepts and practices in core networks, just as applications and services now are logically separated from the transport and access networks. And both those trends illustrate where value is being created within the internet  ecosystem.

Today, owners of networks are virtualizing. Eventually, it is likely to be possible to federate across networks, for that very reason. So essentially, the commoditizing of physical infrastructure will increase.

“Virtualization implies that the control of services and networks can be accomplished outside the physical networks,” and separately from the bit transport layer, argues  Rick Talbot, Current Analysis principal analyst.

That applies initially to any single network, but eventually is likely to extend across multiple federated networks as well, where control lies outside any single transport network.


Network functions virtualization, for example, replaces embedded software in physical network elements across a range of appliances (firewall, packet core, broadband network gateways, customer premises equipment, network address translation, session border controller, provider edge routers), for example.  

“It is only a matter of time before most, if not all, of the specialized network elements and appliances are replaced with--or supplemented by--virtualized instances of themselves,” says Talbot.

There are analogies to this process of core network virtualization. Consider the generation of economic value within the internet ecosystem.


At the moment, perhaps 15 percent of value (measured as firm revenue) is claimed by internet access providers. In the future, that is likely to fall, if mostly because the rest of the ecosystem will have far-higher rates of growth.
To be sure, service providers want and need networks that are more flexible, cost less and also are more capable. Virtualization is one way to achieve those goals. But virtualization also tends to lay bare the growing “dumb pipe” role of networks. Even when networks are “smart,” that smartness is abstracted from the physical network.



At the same time, the separation of logical and physical functions of networks is happening both within the core transport network, within data centers, in access networks and consumer devices.

The obvious analogy is the separation of application supply and network access that is fundamental to all internet protocol networks: any lawful application can be reached and used by end users over an IP network so long as they have authorization to use the apps, irrespective of the IP network used for access.

Where access to communication services once was highly vertically integrated, it now is highly disaggregated. Today, most services and applications used by most consumers are owned by third parties, not the access service providers.

To be sure, access providers (cable, telco, satellite, fixed wireless) have revenue models based on a mix of owned applications (voice, messaging, video entertainment) and “dumb pipe” (internet access, both fixed and mobile) services.

But it is hard to avoid the notion that, over time, value is shifting away from the supply of physical connections to networks (access) and towards “over the top” applications and devices. That does not preclude access providers creating or owning OTT applications themselves.

That, in fact, is what Sling and DirecTV Now are all about; or why Verizon owns automobile application assets.

In a study by AT Kearney of the U.K. internet ecosystem, analysts estimated that internet access accounts for 35 percent of the U.K. Internet value chain revenues (16 percent of the total Internet ecosystem). Excluding the value of e-commerce, the internet ecosystem in 2010 already had apps and services claiming 65 percent of value.


The point is that virtualization is another example of the way networks are changing: as services and features have been abstracted from the network, now even the control and management of networks is being abstracted from the physical facilities.

Monday, March 12, 2018

Data Centers Drive 100G in 2017

Data center adoption of 100 Gbps platforms is driving global Ethernet switch sales, says IHS Markit


Worldwide Ethernet switch revenue grew twopercent sequentially in the fourth quarter of 2017 to $6.7 billion.


For the full-year 2017, revenue rose 8 percent, reaching nearly $25 billion—the strongest growth in seven years, says Matthias Machowinski, IHS Markit senior research director.


“The transition to 25/100GE architectures in the data center is in full swing, driving strong gains in 25GE and 100GE, while in turn bringing down the 40GE segment, which had its first annual decline,” Machowinski said.


100GE Port shipments of 100GigE grew 400 percent year-over-year, and reaching over four million ports in 2017, he said.


Growth in the North American Ethernet switch market was above five percent. Asia Pacific remains the top growth region, primarily due to China.


Cisco’s revenue declined two percent year-over-year, while Huawei grew its revenue 24 percent HPE (Aruba) grew 13 percent.




OTT Versus Linear Video is Becoming a Global Battlefield

By now, nobody is surprised to hear that linear video subscriptions continue to drop or that over the top subscriptions are growing. Perhaps the bigger story is the globalization of the business. Netflix now is a global content supplier, while most other providers operate mostly in a single country, or a small number of countries.

So while it still makes sense to track how U.S. service providers are doing, compared to U.S. competitors, the battle has become global, and Netflix arguably is the leader, in that regard.

Related image

In aggregate, there are more U.S. paid streaming accounts than linear accounts in service.

Netflix has some 55 million U.S. accounts, while Amazon Prime has some 90 million subscribers. All the largest linear video providers together have about 92.2 million accounts.

Total revenue is another story, as monthly subscription revenue earned by a linear account can be an order of magnitude greater than the revenue from any single OTT streaming account.

In 2017, for example, the major U.S. providers lost about 1.5 million accounts, up from some 760,000 in 2016, according to Leichtman Research Group.

The big swing was that streaming services owned by the linear providers gained 1.5 million accounts, nearly the amount lost by the two satellite services.

The biggest six cable companies now have about 48.1 million video accounts. Satellite TV services claim 31.5 million subscribers (including DirecTV, owned by AT&T).

The largest three fixed network telephone providers have 9.2 million subscribers (nearly all provided by AT&T and Verizon).

The top OTT services have about 3.4 million subscribers.

Pay-TV Providers
Subscribers at
End of 4Q 2017
Net Adds in
2017
Cable Companies


Comcast
22,357,000
(151,000)
Charter
16,997,000
(239,000)
Altice
3,405,500
(129,000)
Mediacom
821,000
(14,000)
Cable ONE*
283,001
(37,245)
Other major private companies**
4,200,000
(90,000)
Total Top Cable
48,063,501
(660,245)



Satellite Services (DBS)


DIRECTV
20,458,000
(554,000)
DISH TV^
11,030,000
(995,000)
Total DBS
31,488,000
(1,549,000)



Phone Companies


Verizon FiOS
4,619,000
(75,000)
AT&T U-verse
3,657,000
(624,000)
Frontier
961,000
(184,000)
Total Top Phone
9,237,000
(883,000)



Internet-Delivered


Sling TV^^
2,212,000
711,000
DIRECTV NOW
1,155,000
888,000
Total Top Internet-Delivered
3,367,000
1,599,000



Total Top Providers
92,155,501
(1,493,245)


Netflix Wants Massive Disruption of Every Internet Access Market in the World

The thing about value chains is that, lower in the stack or earlier in the chain, each participant’s revenue is a “cost” to other participants. That explains, as much as anything, industry segment views on network neutrality, inter-segment competition, regulation and antitrust action, irrespective of the public policy goals any particular policy represents.

Be clear, that does not mean there are only naked commercial interests involved. There are legitimate issues of public policy. But for every element of public policy, there are corresponding financial interests.

Example: Netflix views on disruption of access markets. Universal internet access availability and massive internet access cost declines are preconditions for Netflix and its business model.

“Reliance Jio has been a transformational network in India and has brought down data cost massively,” says Netflix CEO Reed Hastings. “There are other people around the world having that kind of impact and can get people to use internet more.”

“We hope someone would do a Reliance Jio in every other country,” says Hastings.

It would be exceptionally hard to find any reasonable objection to the policy notion that every citizen and consumer should have access to the internet, at prices they can afford.

At the same time, it is incontestable that the widespread availability of cheap internet access is a precondition for the business models of every application provider and device manufacturer, while high-quality (fast, reliable, low latency) is a precondition for some business models.

Since application providers sit “at the top of the stack,” every other segment’s revenue streams are “costs” for the application layer providers.

That is true even when the app provider has vertically integrated all or most of the value chain (silicon, processors, optical fiber, communication networks and elements, network management and operating system software, data centers, end user devices, personnel, marketing platforms and fulfillment mechanisms.

That explains why retail telecom service providers care about the cost of servers, switches, construction, optical cable, management and billing software, end user devices and software used to create and deliver the actual end user applications, union contracts and so forth.

All those are cost inputs to the retail price. So for Netflix, universally available cheap internet access is a fundamental requirement for people deriving value from Netflix services. Without internet access, the product has no value.

So it is in Netflix business interests for the rest of the value chain to create and support universal and low-cost internet access. The same is true for Google, Facebook and all other app providers.

If you want to know why industry segments fight over network neutrality, it is because their business interests are at stake. The net neutrality debate is not only about business interests of particular firms and value chain segments. But it explains the positions those interests take.

Sunday, March 11, 2018

Ultra-Low Latency is the Defining Characteristic of Next Era of Networks and Business Models

Eras now change so fast in the computing space that it is hard to come up with nomenclature that makes sense. It used to be easy.

In terms of platforms, we moved from mainframe to minicomputers to personal computers (and client-server) to mobile or ambient computing. Some might say the next evolution is “pervasive computing” based on widespread internet of things adoption.

In the area of core technology we moved from vacuum tubes to transistors to integrated circuits to microprocessors, graphics processors and now seem headed for an era where artificial intelligence becomes the driver.

In terms of apps, we have moved from enterprise back office to desktop productivity to internet apps on smartphones. It is the recent era that seems most diverse. Some people might say we are now in the age of mobile computing. Others might say it is more about “internet computing” or “cloud computing” or “social computing.”

Recent “ages” or “eras” change so fast we probably are not talking about ages or eras at all. Within a span of seven years, for example, we might be said to have moved through the ages of web, to search, to social, to cloud, to mobile.

Obviously we are too granular. An “age” or “era” should not be so short as to last only a year or two. That is a buzzword or a theme, certainly, but not an “age or era.”

At the risk of contributing to the confusion, we might be on the cusp of an era of ultra-low-latency networks designed to support new applications requiring such ultra-low latency, one of the defining performance characteristics of 5G.

But there are other big changes. A new edge computing infrastructure is believed necessary to support the ultra-low-latency apps for which 5G will provide the access. Inside the core network, new methods of creating virtual private networks (including network slicing) might also sometimes play a role in creating low-latency networks.

Most likely, the coming era will not be called anything like “ultra-low latency.” It simply is not catchy enough. “Era of IoT” or “ambient computing,” while not especially easy  to grasp, might make more sense.

Still, the key change is latency performance, though most casual observers will point to the huge increases in bandwidth as the defining characteristic. But when Nokia Bell Labs consultants talk about “creating time,” they are speaking broadly about latency performance, not bandwidth.

And to the extent that 5G revenue upside comes from internet of things apps (sensor connections), much of the clear upside comes in the ultra-low-latency areas, as much of the IoT sensor access requirement might be met by existing 4G or fixed connections (Wi-Fi, for example).

Many sensors will have essentially zero latency dependence, as they might report data only once a day or at specified other intervals, but with no particular dependence on latency performance.

Other applications, including autonomous vehicles, virtual or augmented reality, remote surgery or instant startup of 8K video content, might well require latency below 10 milliseconds.


The implications are fairly clear: if in fact the next big wave of revenue opportunity for service providers are various internet of things (sensor) applications--not humans using smartphones--the key area where 5G is unique is ultra-low latency, since 5G is specifically built to deliver ultra-low latency.

So it is not bandwidth, as such, that is the clear differentiation between pre-5G and 5G and future networks. Latency performance is where the difference exists, not bandwidth, though the increases will be significant.

Even advanced 4G networks, Wi-Fi and other networks will routinely operate in the hundreds of megabits per second to gigabit ranges, and most apps will not routinely require that much bandwidth.

It will be some time until we are able to come up with a universally-accepted nomenclature for the coming era that includes 5G and edge computing. As has been the case in the past, we will be able to characterize the eras at the component level, the computing model, the devices used or by lead applications.

And, cumbersome though it might be, latency will figure in nearly every characterization, and possibly in all descriptors. Low latency really is the key, where it comes to the business model, the key changes in architecture or revenue model changes.

Thursday, March 8, 2018

U.S. Internet Access Prices Actually Have Not Increased Over 2 Decades

As the old adage implies, people are “entitled to their own opinions, but not their own facts.”

So it is that historical evidence for consumer internet access “price gouging” is virtually impossible to find in the U.S. market, over the past two decades, when internet service providers have--for all but three years--been able to charge whatever the markets would bear.

In fact, improvements in product quality (bandwidth, for example) have been made almost at Moore’s Law rates. And prices broadly have dropped, in absolute terms as well as relative terms, without considering inflation adjusted prices or product changes.

Impressionistically, recall only that some of us who were buying “broadband” internet access before the turn of the century were paying hundreds of dollars a month for 756 kbps service. Today we are able to buy 100-Mbps service for $50 to $70 a month, on a standalone basis.

In fact, because the internet access actually is part of a bundle, the imputed cost of 100-Mbps fixed network internet access might be less than $50 a month.

And since many of us buy mobile bundles as well, the cost of 15 Gbytes of monthly usage, per device, might be about $30 a month, per device. That is less money than we used to spend, for smaller usage buckets.

People with shorter experience of internet access prices might not understand the price declines.

Consumers who do not consider the changes in volume or the actual discounts on the products they actually buy might not be so aware of the price changes.

But assume one argues typical posted retail prices have climbed over the last decade or so. Ignore for the moment the fact that the “product” itself has changed, climbing from less than 1 Mbps to 100 Mbps to 300 Mbps for standard products.

Even there, one can argue that the increase in retail prices has been less than the overall level of consumer prices. In other words, internet access posted retail prices have increased less than the general level of prices.

The argument that ISPs are oligopolists whose behavior “harms” consumers because prices are higher than they ought to be, and service quality lower than it ought to be, is not borne out by the historical record.

According to the U.S. Bureau of Labor Statistics, for example, between 2009 and 2017, the producer price index (PPI) for “telecommunication, cable, and internet access services” shows that prices actually dropped, and dropped fastest in 2017.

What makes this data even more noteworthy is that it includes “cable TV” subscription services, which everyone agrees climb every year, at rates above the background rate of inflation generally.

So if cable TV prices rise every year, the only explanation for declining prices for the whole basket (telecom, internet access and video) is that telecom and internet access prices dropped even more than cable TV prices rose.

Producer Price Index
Telecom, Cable TV, Internet Access

Likewise, as consumer prices have increased (from 2007 to 2012, for example), the cost of internet access increased less than the overall CPI, even as product quality (measured in terms of speed) increased.

In earlier periods, such as 2000 to 2009, we see the same trend: U.S. internet access prices dropped about 80 percent, according to the Economist.



Looking at cost per bit, since 1999, cost has dropped by two orders of magnitude (100 fold).


Historically speaking, U.S. retail internet access prices have dropped, not risen, over the past two decades, on a cost-per-bit basis; compared to the consumer price index or other inflation-adjusted terms.

Some of the price declines are hidden, as consumers have shifted to bundled products, not standalone products, where prices are effectively discounted, even if standalone retail prices grow.

Finally, the product itself has changed. We now routinely buy services that are two orders of magnitude better, at prices four to six times cheaper.

Many Winners and Losers from Generative AI

Perhaps there is no contradiction between low historical total factor annual productivity gains and high expected generative artificial inte...