Global Bandwidth Demand Generated by a Relative Handful of Data Centers

Some might recall a time when international capacity existed mostly to transport voice calls. In 1980, for example, tier-one telcos were the owners of the transmission assets and trans-Atlantic capacity could easily be measured in kilobits per second up to 1.5 megabits per second. 

These days, capacity on global routes is measured in terabytes. In 1980, most of the demand was generated by people making long distance calls. Today, internet apps and video drive most of the global demand for capacity.

source: AI Trends 

A small handful of firms--Google, Facebook, Amazon, and Microsoft—drive most of the demand for international capacity. In 2020, those four firms accounted for 66 percent of all used international capacity. So content services and applications now drive demand for global bandwidth.

About 45 percent of such hyperscale data centers are located in the United States, about eight percent in China, seven percent in Japan, for example. 

source: Synergy Research

Beyond that, the owners of facilities are mostly a handful of content and application providers, not telcos or connectivity service providers. 

Annual demand still grows close to 50 percent, especially on routes connecting Africa and Asia. On other routes, annual growth ranges between 30 percent and 40 percent, according to Telegeography. 

source: Telegeography 

One might think most of that capacity is used to deliver content and data to end users of the internet. Actually, most of the bandwidth connects hyperscale data centers operated by the largest app firms. 

source: Vox.com, Cisco 

source: Cisco 

Among many other changes to the global telecom business caused by a shift to internet, mobile and cloud communications, global bandwidth patterns have changed. In the past, long haul traffic originated and terminated at central offices. 

Now, with the dominance of mobile-consumed content and apps, using the internet, cloud computing and web browsers or apps, traffic originates and terminates heavily at hyperscale data centers.

source: Cisco

Asia-Pacific has been the fastest growing region in terms of hyperscale data center location and will continue to grow more rapidly over the next five years, although North America accounted for 43 percent of hyperscale data centers in 2020, says Cisco.

The point is that global capacity demand has shifted dramatically: from telcos to app and content providers; from voice to internet traffic; from voice to video; from end users to data center interconnection.

Cost Certainty Has High Value for Consumers

Enticing consumers to upgrade to a next-generation network (fixed or mobile) might not be difficult for early adopters, who care less about cost than “being first” to get the new technology. Most consumers need other justifications, though. 

Some perceived problem needs to be addressed; some new value has to be added; new capabilities or use cases have to be made clear, before mainstream consumers will feel they must upgrade. 

Sometimes network performance does the trick; sometimes not. Where it is available, 5G can claim to offer faster internet access speeds. Depending on which spectrum options are available, the incremental change might be almost non-existent, slight or highly differentiated. 

Based on past experience, mobile customers know a bit of trial and error is possible or required. 

Right now, U.S. mobile operators often take a different tack. Rather than hanging the purchase decision entirely on 5G network performance or availability, they shift the value proposition to cost certainty. 

In other words, 5G is offered as a feature bundled with unlimited usage plans. But the key value is “cost certainty.” Customers know what the recurring cost will be, and need not worry about additional usage or overage charges. 

Even when an unlimited usage plan costs more, there is enough additional value in such plans to drive adoption, so valued is cost certainty. 

Predictability also is valued in terms of product consistency, others will note. That seems less the case for unlimited mobile data usage plans. What really matters is cost certainty. 

Consumers tend to prefer predictable cost services because it allows them to budget for the  recurring payments. People do not like surprises, where it comes to recurring purchase costs. 

There is considerable evidence of consumer preferences for subscription over per-use pricing. 

During the 1970s, the U.S. Bell system telephone companies started offering customers a choice between the the traditional flat rate option, which might cost $7.50 per month, and allow unlimited local calling, and of a measured rate option, which might cost $5.00 per month, allow for 50 calls at no extra charge, and then cost $0.05 per call. 

In the numerous trials that were carried out, the flat rate option was usually selected by over half of the customers who were making fewer local calls than the 50 covered by the measured rate basic charge, even though they clearly would have benefited from per-use pricing. 

These results are documented in Cosgrove and Linhart (1979), Garfinkel and Linhart (1979) and Garfinkel and Linhart (1980). 

The same sort of issue arose in the 1980s and 1990s around videocassette and DVD movie rental fees. People were charged a late fee when cassettes were returned “late.” Some argue that early Netflix, which mailed DVDs out to consumers, succeeded in part because there was no danger of late fee charges . 

In fact, the Netflix subscription model was designed to eliminate late fees.  

Similar results were seen when long-distance calling plans were similarly varied in the 1980s.Consumers could pay by use or buy flat-rate calling plans. Many customers  paid for plans that provided more calling than they actually used, as documented by Mitchell and Vogelsang (1991).

In the internet era, consumers often buy usage plans that offer more usage than they expect, simply so customers are assured of the recurring cost.

There are three main reasons that probably lead consumers to prefer flat-rate pricing, and they were recognized a long time ago in research by Cosgrave and Linhart (1979), Garfinkel and Linhart (1979) and Garfinkel and Linhart (1980).

Fixed price plans provide protection against sudden large bills. Customers also typically overestimate how much they use a service, leading them to buy plans that they actually do not need.

There is a convenience issue as well. In a per-use situation, consumers seem to worry about whether each call is worth the money it costs. That limits usage, as a study by Garfinkel Garfinkel and Linhart (1980) found.

 A flat-rate plan allows them not to worry about the cost of each call.

The point is that the reasons consumers prefer flat rate service plans are clear. They value such plans enough to pay more than they might under usage-based plans. But the cost certainty has value. 

In the case of 5G unlimited usage plans, that behavior will be clearly seen. The value might often be less the “unlimited usage” than the “certainty of cost.”

No Digital Transformation Without Computing Cost Reductions

Economist William Nordhaus has estimated that computing costs have fallen dramatically since the 1940s, improving by an order of magnitude every four years in the 1980s and 1990s, for example. You can safely argue that this vast reduction in cost underpins the shift to affordable digital transformation of all types, not to mention the broader shift from physical to digital processes. 

source: Nordhaus 

Scale Matters for Cloud Computing Economics

Scale matters where it comes to many computing use cases. As useful as hosted voice services are, total cost of ownership is higher for large enterprises than an “owned infrastructure” approach, primarily because costs are directly related to the number of lines supported. 

Some companies with high reliance on computing facilities, such as software firms, will often find they can cut costs dramatically by shifting computing operations back to their own facilities, and off computing as a service platforms, partners at Andreessen Horowitz argue. 

That appears to be especially true for public software companies, where the computing infrastructure is an unusually-high percentage of total costs, Andreessen Horowitz argues. 

“When you factor in the impact to market cap in addition to near term savings, scaling companies can justify nearly any level of work that will help keep cloud costs low,” say Sarah Wang and Martin Casado, Andreessen Horowitz partners. 

“A billion-dollar private software company told us that their public cloud spend amounted to 81 percent of cost of revenue,” they say. Also, “cloud spend ranging from 75 to 80 percent of cost of revenue was common among software companies.”

Citing the example of Dropbox, the partners note that “when the company embarked on its infrastructure optimization initiative in 2016, they saved nearly $75 million over two years by shifting the majority of their workloads from public cloud to ‘lower cost, custom-built infrastructure in co-location facilities’ directly leased and operated by Dropbox,” 

Dropbox gross margins increased from 33% to 67%  from 2015 to 2017, which they noted was “primarily” due to Infrastructure optimization and an increase in revenue. 

Thomas Dullien, former Google engineer and co-founder of cloud computing optimization company Optimyze, estimates that repatriating $100 million of annual public cloud spend can often result in nearly $50 million in annual total cost of ownership from server racks, real estate, cooling, network and engineering costs, they note.

Other sources estimate savings ranging from 33 percent to 50 percent, they add. “A director of engineering at a large consumer internet company found that public cloud list prices can be 10 to 12 times the cost of running one’s own data centers, the partners say. 

Smaller firms, earlier in their growth cycles, almost always benefit from cloud computing, rather than building their own computing infrastructures. But that can reverse when firms grow larger and growth slows. 

“It’s becoming evident that while cloud clearly delivers on its promise early on in a company’s journey, the pressure it puts on margins can start to outweigh the benefits, as a company scales and growth slows. Because this shift happens later in a company’s life, it is difficult to reverse,” say Sarah Wang and Martin Casado, Andreessen Horowitz partners. 

source: Synergy Research 

“You’re crazy if you don’t start in the cloud; you’re crazy if you stay on it,” the paradox becomes. 

The issue is how cloud spending versus owned facilities plays out over time for firms in other lines of business that might not have computing cost as such a large driver of total cost of revenue. 

For most connectivity providers, computing cost is not likely a huge driver of total cost of revenue. Information technology costs vary by industry from 1.4 percent in manufacturing  to 11.4 percent in the financial services industry, for example. 

 

source: Computer Economics 

Others estimate information technology spending in about the same ranges, though spending varies by firm even within each industry. Some might argue maintenance levels of spending could range about two percent of revenues, while growth spending might range up to perhaps six percent of revenue. 

Perhaps most firms spend between two and three percent of revenue on all IT. 

 

source: Statista 

Few industries have computing costs as such a large driver of cost as do software firms. Few firms in most industries can improve equity valuations or profits as much as software firms can by optimizing computing cost. 

For most firms in most industries, the balance of owned versus “computing on demand” spending might not be crucial in that regard. 

In principle, if the same savings software firms obtain by optimizing computing spend were to hold universally, most firms in most industries could save relatively little by optimising computing choices. Saving 33 percent, if possible, on an item that represents two percent to three percent of total cost of revenue is helpful, but arguably not all that significant as a driver of cost and profit.

Vertical Integration Versus Portfolio of Assets: Which Works Better?

“Vertical integration” can be problematic for any non-connectivity asset held inside a telco entity. Ask AT&T or Singtel. 

AT&T’s purchase of Time Warner was an effort to vertically integrate into content. The goal, says AT&T CEO John Stankey, was to “help our domestic connectivity business.” As AT&T now sees matters, the streaming content business must be built globally, and not restricted to a single country, which is primarily AT&T’s connectivity business base. 

That--at least as AT&T now positions the matter--requires a globally-focused business unencumbered by considerations of the domestic market. So AT&T wants to combine its Warner Media assets with Discovery, creating a larger new entity in which it owns 70 percent of the equity. 

That will remove Warner from AT&T’s consolidated results. Needless to say, valuation of such a standalone business should exceed the valuation if the asset were embedded within a telco organization. 

Among the other possible ways to view such assets is the ability to produce free cash flow, revenue growth and profit margins for the mother company, even if not fully consolidated within the telco itself. 

In fact, valuation might be higher precisely because a “pure play” content or streaming asset is valued as other similar assets are. 

Singtel, Southeast Asia's largest telecoms operator, reported that annual net profit halved to S$554 million ($418 million), the lowest net profit in at least two decades. That performance by a leader in exploring additional lines of business illustrates three related issues.

The first issue is the exhaustion of legacy connectivity services markets. The second issue is the traditional difficulty or entering new markets, either in other parts of the ecosystem or in content or applications. 

The third issue is the valuation penalty any successful “up the stack” asset (content, platform, application) has when buried within the telco organization. 

Singtel has been trying to diversify for years, and has been a leader in exploring growth “up the stack” in the application layer, and beyond connectivity. So the inability to reap profit rewards is troubling. 

But some investments such as those in digital marketer Amobee and cyber-security firm Trustwave yielded weaker-than-expected returns, observers note. That is a recurring story for telcos who have tried for many decades to broaden their revenue bases in adjacent areas such as software or computing services. 

The next moves might pair infrastructure asset sales to fund investments in other growth areas such as financial services or gaming. 

At least part of the problem is valuation of telco infrastructure assets. Singtel notes that its infrastructure assets do not provide a valuation boost, compared to other suppliers that own fewer network assets. 

The other issue is that Singtel executives believe ownership of towers, satellites, subsea cables and data centers has not boosted Singtel’s valuation, compared to peers who own less of such assets. The expectation is that selling some of those infrastructure elements will free up capital to deploy in other growth areas. 

Weakness in mobile services revenue and market share is among the current issues. Mobile service revenue dropped 19 percent; blended average revenue per user fell 18.5 percent and subscriptions fell 3.6 percent over the last year, for example. 

That is not a unique problem. Globally, other service providers face low growth rates and falling ARPU. Saturation of mobile services--the industry growth driver for decades--is part of the problem. 

Beyond that, the typical telco must replace half of existing revenue every decade or so. We have seen that in fixed network voice services, mobile voice, long distance revenue and fixed network services generally. We have seen it in text messaging as well. 

source: IP Carrier

source: FCC  

That might seem hyperbole, but is a demonstrable fact. Globally, that means telcos have to generate about $400 billion in new revenue just to replace what they will lose over the next decade.  

Singtel’s issues really are the same issues every connectivity service provider eventually will face. The issue is how to create huge new revenue streams, outside the connectivity core. Those who argue that amount of growth can happen within the connectivity core, it seems to me, have the burden of proof.

Vertical integration might not be the only way--perhaps not the best way--to do so. Any sufficiently large asset outside the connectivity core will get a higher valuation if outside the telco, operating as a "pure play."

That might not help connectivity business revenue and profit matgin directly, but it arguably is a better way to grow the overall business.

Where AT&T, T-Mobile, Verizon See Fixed Wireless Value

“I don't expect that fiber will ever be the solution for all of the ILEC footprint,” said John Stankey, AT&T CEO, commenting on fixed wireless as a replacement. “We're not as robust in our point of view on what fixed broadband can do in urban and highly attractive suburban areas.”

“But what we do believe is that fixed wireless plays a role in other parts of our footprint,” he said, especially in rural areas. “That's a really nice replacement for some percentage of the data customers that are out there, and we'll continue to pursue that,” he added. 

That more-cautious perspective on fixed wireless, compared to Verizon and T-Mobile, grows from the respective positioning each firm has in the market. 

AT&T has the largest footprint of U.S. households among U.S. telcos, and therefore has the most to lose, and the least to gain, from new investments within the more-rural parts of its fixed network territory. 

Verizon has a relatively small footprint of U.S. homes, and therefore has much more to gain from fixed wireless using its mobile network, out of region.

T-Mobile has zero market share in home broadband, and clearly  has the most to gain, in terms of additional revenue and market share in that business.

The Vertical Integration Downside: Lower Valuation

Singtel, Southeast Asia's largest telecoms operator, reported that annual net profit halved to S$554 million ($418 million), the lowest net profit in at least two decades. That performance by a leader in exploring additional lines of business illustrates three related strategic issues that eventually will be faced by every connectivity service provider serving the mass market.

The first issue is the exhaustion of legacy connectivity services markets. At some point, every customers that wants service will be buying it.

The second issue is the traditional difficulty or entering new markets, either in other parts of the ecosystem or in content or applications. Telcos have had a very-difficult time sustainably creating new roles for themselves elsewhere in the value system.

The third issue is the valuation penalty any successful “up the stack” asset (content, platform, application) has when buried within the telco organization. The same asset--inside a telco--earns a less robust valuation than that same asset, outside the telco organization.

That raises the issue of "how" a connectivity provider should "own" assets elsewhere in the value chain. Vertical integration actually seems to penalize the value of assets. Which suggests some strategy of ownership "outside" the connectivity organization.

Singtel has been trying to diversify for years, and has been a leader in exploring growth “up the stack” in the application layer, and beyond connectivity. So the inability to reap profit rewards is troubling. 

But some investments such as those in digital marketer Amobee and cyber-security firm Trustwave yielded weaker-than-expected returns, observers note. That is a recurring story for telcos who have tried for many decades to broaden their revenue bases in adjacent areas such as software or computing services. 

The next moves might pair infrastructure asset sales to fund investments in other growth areas such as financial services or gaming. 

At least part of the problem is valuation of telco infrastructure assets. Singtel notes that its infrastructure assets do not provide a valuation boost, compared to other suppliers that own fewer network assets. 

The other issue is that Singtel executives believe ownership of towers, satellites, subsea cables and data centers has not boosted Singtel’s valuation, compared to peers who own less of such assets. The expectation is that selling some of those infrastructure elements will free up capital to deploy in other growth areas. 

Weakness in mobile services revenue and market share is among the current issues. Mobile service revenue dropped 19 percent; blended average revenue per user fell 18.5 percent and subscriptions fell 3.6 percent over the last year, for example. 

That is not a unique problem. Globally, other service providers face low growth rates and falling ARPU. Saturation of mobile services--the industry growth driver for decades--is part of the problem. 

Beyond that, the typical telco must replace half of existing revenue every decade or so. We have seen that in fixed network voice services, mobile voice, long distance revenue and fixed network services generally. We have seen it in text messaging as well. 

source: IP Carrier

source: FCC  

That might seem hyperbole, but is a demonstrable fact. Globally, that means telcos have to generate about $400 billion in new revenue just to replace what they will lose over the next decade.  

Singtel’s issues really are the same issues every connectivity service provider eventually will face.