Beyond "Public" Digital Infra

For some, digital infrastructure means data centers, cell towers and optical fiber networks. For others the category includes software, computing hardware and devices, artificial intelligence and other  information technology. For investors, operating companies, asset owners and many others in the category, the definitions matter. 

source: IDC 

For investors, it changes the classes of assets that are investable. For operators the definitions shape business strategies. For asset owners, broader definitions change the scope of potential buyers. For everyone, the definitions can affect equity or asset values and multiples. 

source: YTD2525 

Likewise, broader or narrower definitions shape firm strategies for all others in the ecosystem: who “customers” are, how marketing and sales efforts are structured, what product development has to happen. 

For most firms, a practical definition will be narrower. But it seems inevitable that with the emergence of artificial intelligence as a trend at an inflection point, that the category will see practical changes, with more asset types emerging as part of the digital infra marketplace. 

Not to be ignored either is the amount of competition for the narrower class of assets, higher valuation multiples that put pressure on upside profits and the sheer emergence of new asset classes closely related to traditional digital infra. Also, existing asset classes could emerge in new form as part of digital infra markets. 

Historically, “public” assets such as communication networks or data centers have been distinct from “private” assets such as computing hardware or operating systems; end user applications and software. 

We should see a growing trend of greater inclusion of other asset classes beyond “public” infra.

So ISPs Want a Massive Shift to "Collect Calling" for Content Delivery?

Virtually no observers of the Western Europe connectivity business would challenge the notion that the business model is challenged. Typical complaints include consumer resistance to pay for connectivity; competition that is ruinous and government policies that incentivize competition more than investment. 

Infrastructure financing was a significant theme of connectivity provider and regulator talks at MWC Barcelona, as you would expect. 

The argument normally is phrased as a matter of “fairness,” a handful of hyperscale app providers “causing” more than half of all data usage by an ISP’s own customers. But that matter also can be looked at through the lens of the way access providers always have compensated each other for traffic imbalances. 

Keep in mind it always has been the networks that compensate each other. Think back to the say connectivity networks used to work in the voice era. Calls necessary involved at least two parties and locations. 

Resources were consumed on each end of the connection. In most cases, the charging model was “calling party pays,” with a healthy dose of “both parties earn revenue.” Though the calling party paid for long distance charges, the terminating party also was paid for use of the network. 

In the era of internet content delivery, one has to ask “who is the calling party?” Is it the ISP customer who asks for delivery of content? Or is it the content owner who responds to a consumer’s or customer’s request? Is the “request” a different session from the content delivery? Or is the content delivery a response to the calling party’s request? 

In other words, are we talking about “fairness” or “calling party pays?” And what is the traffic source and sink? 

soruce: flylib

It might seem obvious that a content provider is the “source” while the requesting subscriber or user is the “sink.” But that transaction or session only happens because a customer or user of one ISP has asked for that content from an app provider on another network.  

In other words, which party placed the call? And which charging method should apply? It’s complicated because there is no internet access equivalent to the call “minute of use.” There is no rating system in place we can use to track and then allocate network usage charges when sessions and data flow between ISP domains. 

In an era where networks mostly are used for content delivery, asymmetrical traffic demand is assumed.

Think back to the era when network traffic was mostly voice. By definition, traffic flows were nearly symmetrical. No matter which party called, and which party was called, capacity resources were virtually identical. It did not matter which party talked, and which party listened. Network resource consumption was essentially identical on each end. 

In principle, there are three ways to account for usage by customers on two different networks. The calling party can pay; both parties can be paid or the receiving party can pay. The only other pattern is that a third-party sponsor pays (toll free calling, for example). 

In the fixed networks business the general pattern has been calling party pays. In the mobile business both parties have paid, in the form of usage allowances being depleted on both ends. And calling parties have paid for toll charges. 

In the internet access business, the general pattern has been both parties pay, in the form of usage allowances that are paid for and consumed. Proposals to tax a few hyperscale app providers retain the “both parties pay” scheme, but also add a new layer of “some app providers also pay.” 

Many analogies could apply. Placing a long distance call might be one analogy. Even when resources are used on both ends, it is the caller who pays the charge. Applied to internet content delivery, it is the ISP customer who invokes the data delivery who might be viewed as the entity that pays. 

The other approach--receiving party pays--only applies for special circumstances similar to a “collect call,” where the called party agrees to pay the charges normally paid by the party placing the call. 

The proposal to tax a few hyperscalers is equivalent to collect-call billing procedures, where the party initiating the call is not billed, but the party at the other end is billed instead. One might say that proposals to tax a few hyperscalers are a form of billing similar to the “collect call,” where the calling party is not billed. Instead, charges are borne by the called party. 

That is not to say such practices are inherently flawed. It simply is to note that the “both parties pay” is the most-widely-used pattern for mobility networks, even when calling party pays also is used and collect calling is conceivable. 

Even if we agree that calling party pays is reasonable, who is the calling party? Connectivity provider calls for payments by a few hyperscalers are more akin to “collect calling.” 

That might seem workable to many. But the additional charges borne by the entity paying for the collect call still must account for such charges in their business models. And those higher costs will be borne by business partners, retail customers or users. 

Gains by ISPs will be offset by losses by content or app providers. So higher costs will accrue to advertisers, sponsors, retailers and users and subscribers of such services. There is no “free lunch.”

Is Australia's Home Broadband Market Saturated?

Is home broadband in Australia close to complete saturation? If so, that implies the home broadband business in Australia now is virtually a zero-sum game. What one internet service provider gains is offset by an equivalent loss by some other contestant.

In such markets, it is not only tough for internet service providers to grow revenue by adding subscribers, but the revenue growth model has to be based on the ability to increase the revenue each provider gets from its customers.

Typically, that involves getting customers to buy more-expensive plans, almost always based primarily on higher speeds, though the value of bundled features often is hard to separate from higher-speed access.

New data from the Australian wholesale connectivity provider NBN shows a slight dip in wholesale accounts supported by the network. For some observers, that is the key statistic. It might indicate that demand for home broadband is nearly saturated, or is saturated. 

Assume the NBN represents the whole home broadband universe, representing 8.7 million accounts. There are some satellite accounts, but too few to affect the analysis (fewer than 21,000 total accounts nationally). 

If there were some 10.9 million dwelling units in Australia in 2021, and about 178,000 new units are added each year, then in 2023 there are about 11.3 million dwellings in early 2023. 

source: id.com 

That implies a take rate of about 77 percent. The slight dip in current subscriptions suggests that demand is about as high as it can get. 

source: ACCC 

Some see a market share shift from larger internet service providers using the NBN (Telstra, TPG, Optus) to smaller ISPs, a trend that was consistent in 2022 and 2021. 

It appears that the smaller ISPs are succeeding in the ways smaller providers often do. Anna Brakey, commissioner of the Australian Competition and Consumer Commission. “Some smaller providers are offering consumers different options to meet their specific needs, such as tailored plans and discounted pricing options, network performance graphs, Australia-only call centers and gamer-optimized plans,” she said. 

But the main observation might be the drop in total subscriptions in a developed country market, which is unusual.

Choice of TCP/IP Was Fateful

Choices have consequences. When the global connectivity industries decided that TCP/IP was the next-generation network, they also embraced other foundations. Functions now are layered. We use application programming interfaces to communicate between layers, but the layers themselves are disaggregated. 

That means monolithic value chains or vertically-integrated value chains are also disaggregated. One can own and operate an asset at one layer without owning and operating all the other layers. App ownership and creation is separated from delivery, for example. 

That is why “over the top” is essentially the way all apps now are created. Which entity “owns” an app can vary, but development is permissionless. 

In a similar and earlier way, the digitization of all media types has had key consequences. Digital media means digital delivery. And that also means delivery in a permissionless way over layered networks. Where content owners once had to create or own their own delivery networks, now any media type can be accessed by any user (so long as it is lawful in the eyes of a government) without a specific business relationship between content owner and distributor. 

That enables content streaming, which has created traffic imbalances between inbound and outbound data. Only kilobytes need be sent upstream to request delivery of a two-hour movie, whose delivery entails gigabytes. 

And though the magnitude of data transfer is less for other types of content and interactions, the same imbalance exists. It only takes kilobytes to request a page or an object. The delivered content, ranging from websites with auto-start video to streaming audio, requires much more data to be delivered and displayed. 

The public communications networks long have had a way of dealing with imbalances of this type. At the end of a year, carriers true up usage. If one network has landed more traffic than it has sent--and in principle “used” more resources--payments are made by the sending network. 

And that is the logic behind proposals to tax a few hyperscalers for landing traffic on ISP networks. 

Think about the impact on networks in making the switch from analog to digital; linear to on-demand delivery. Content delivery networks always are most efficient when using a broadcast or multicast model where essentially one copy is delivered at the same time to many thousands to millions of viewers or listeners. That allowed one-to-many networks to be built and operated by the content companies (radio stations, TV broadcast stations, cable TV). 

On-demand delivery requires a very different kind of network. Unicast delivery then must be supported. Any-to-any networks require overbuilding capacity, compared to a broadcast network. Where in a linear environment one copy is sent at the same time to many consumers, a unicast network requires sending one copy to one consumer, each requiring consumption of additional network resources. 

Of course, it is complicated. It is the ISP’s own customers who are invoking the remote data. If the product were electricity, water, natural gas, toll road access, landing rights, docking rights, lodging nights or most other retail products, buyers pay for usage. Business-to-business usage winds up--ultimately--paid by end user consumers, even if intermediate costs are borne by other business partners. 

Intermediate funding can come in various other ways. Advertisers or sponsors can defray some costs, those costs being recouped in sales of whatever products advertisers are hawking. 

Some participants can be taxed or subsidies can be applied. Taxes on a few hyperscalers illustrate the former; universal service subsidies represent the latter approach. 

The larger point is that choosing TCP/IP has had business model consequences. Permissionless app development means the financial interests of app owners and network owners can be disaggregated. And, as in any value chain, one participant’s revenue is another participant’s costs. 

Ultimately, all long-term value chain costs are reflected in retail end user prices. But costs and revenues within the value chain always are contentious to some degree. And any long-term increase in producer prices will be reflected in higher consumer prices, reduced output; lower producer profits or other changes in features. 

No matter what the resolution of debates over funding mechanisms, retail consumers will pay, in the form of higher connectivity fees; higher retail product costs or changes in feature sets. 

Producers could see pressure on profit margins if their own capital investment and operating cost parameters are not adjusted. 

But it all goes back to the fateful choice of TCP/IP as the next-generation network. Layered and disaggregated models have consequences.

Private Equity Feels Impact of Higher Real Interest Rates

Exuberance in public markets seems to have been matched in private markets, as ultra-low interest rates shaped the investment climate in all markets. Higher real rates are having the opposite effect, as Adams Street Partners data suggests. 

source: Adams Street Partners

Deals will be smaller and the volume of transcations will drop as a result. Less money raised also means less money invested. That should mean longer runways to exits.

Fixed Wireless Really is Affecting U.S. Home Broadband Share

Fixed wireless has emerged as the clear producer of “new revenue” for 5G networks. At least for the moment, FWA seems to be crimping cable operator home broadband net additions. 

In the third quarter of 2022, for example, Comcast added 14,000 net new subscribers. T-Mobile, using FWA, added 578,000 net new accounts. Verizon’s FWA service added 342,000 net new accounts, according to Leichtman Research Group figures. 

Legitimate questions can be asked about how long that trend can last, as most observers would agree that FWA appeals mostly to customers without significant need for, or desire for, the faster services. In other words, for a significant portion of the market, speeds up to 100 Mbps to 200 Mbps are good enough. 

But that is not the whole market. In the third quarter of 2022, for example, it is possible that a quarter of all customers, perhaps as many as half, only “needed” speeds up to 200 Mbps. Only about 15 percent of households bought service at 1 Gbps or faster. 

source: OpenVault  

At some point, the “lower-speed is good enough” segment will be saturated. At that point, Verizon and T-Mobile will have to do the same thing as Comcast: boost speeds for existing customers and those who want higher speeds. 

Cable executives predictably expect that FWA will not be able to keep up. Verizon and T-Mobile obviously say they disagree, and that they will be able to keep boosting FWA speeds. 

At the moment, Comcast’s strategy seems to acknowledge the new competition, as it no longer says it will grow home broadband revenue by increasing market share or the number of subscriptions, but rather by upgrading speeds to faster tiers that cost more. 

Competition from fiber-to-home providers is the other part of the market dynamic, as more FTTH is being activated by many ISPs, competing with the fastest of cable home broadband speeds. 

For all those reasons, cable’s revenue growth hopes likely hinge on taking greater share in the mobile phone business.

Can ISPs Really Build Ecosystems?

If you have been in the connectivity business long enough, you are used to hearing visions of how connectivity providers can “revolutionize” their businesses by creating  new lines of business, crafting new products and building new revenue models. 

So Telstra CEO Vickie Brady says the role of the operator is to become an “ecosystem builder,” even if that means “not always being in control of the end-to-end solution.” 

There is a clear logic. Many industries based on software and computing resemble ecosystems. Large data centers these days might be likened to ecosystems, where it is not simply servers, but the connections between firms, servers, apps and software and connections to other locations that create value.

The same observation might be made about the Applie iPhone business, which increasingly is an ecosystem of products built around the device. Even an airliner or an electric vehicle might be said to be an ecosystem that creates value only when the extensive wraparound exists.

Planes need many things to become part of an airline operation. Branding, training, reservation systems, airports, maintenance facilities, business alliances, loyalty mechanisms all are necessary. But the airline industry also is part of a larger travel ecosystem including lodging, local transporation, destinations and attractions.

Cars are themselves an ecosystem of parts and systems, but also require fueling stations, maintenance, insurance, road systems, parking, driving instruction or training. Cars are part of a broader transportation infrastructure that includes other transport modes and also shapes where housing and businesses are located.

We can add "ecosystem builder" to the list of stock phrases such as “telcos becoming techcos,” or becoming “platforms,” trying to create app stores, getting into financial services, mainframe computing, system integration, devices or more recently, edge computing. 

Telcos tried to create their own “over the top” voice over IP services to compete with the likes of Skype, their own messaging apps, their own content services (with more success). 

That is not to throw shade at the companies we once knew as “telcos,” but many decades of efforts at reinvention have had modest success. The big problems are changes in customer demand, competition, product substitutes and the chosen architecture of service provider architectures. 

When connectivity service providers chose TCP/IP as their next-generation architecture over asynchronous transfer mode and the whole suite of ISDN-derived standards, they also chose a layered model that not only permits, but encourages, third party app development 

Since functions are logically separated, no business relationship has to exist between a particular app delivered over the IP network and the owner of the access facilities. So “over the top” is an architectural rule, not a term for streaming video. 

The practical effect is to separate app creation from network services. The former no longer requires ownership of the latter, a contract with the latter or the permission of the latter. And while connectivity providers had developed voice and texting, they had no special competence in creating apps for computers or computing devices. 

And these days, that is most of app development. 

At the same time, competition has taken away monopoly-era profits and gross revenue and market share. Customers, meanwhile, prefer mobility services over fixed network voice services, and messaging over short message service. 

All of this challenges the business model. 

To be sure, one might point to growing global services revenue, as more people become mobile subscribers, in particular. But most legacy tier-one service providers have seen flat or declining revenues, challenged profitability, profit margin squeezes and declines in average revenue per account. 

That, one might argue, bolsters Brady’s argument that massive change has to happen. But for the rise of mobility services and the internet, legacy service providers would be in even worse shape than they are. 

So Brady’s call to create new ecosystems with connectivity providers at the center is not untimely. Skeptics will question which entities actually could emerge at the center of new ecosystems of value, or whether additional revenue actually would accrue to connectivity providers in such ecosystems. 

One might, for example, question whether the sale of functions necessarily nets more revenue than the disaggregated elements. Many observers already fear any relegation of access providers to the role of “bit pipes” or “dumb pipes.” Some might say the sale of deconstructed features could help or hurt, in terms of maintaining relevance and value. 

There is gold, there are gold miners and there are many other roles that support gold mining, from hardware to services. What Brady suggests is that connectivity providers could organize or participate in the equivalent of a gold production ecosystem. That is fair enough. 

The issue always is core competence and how additional competencies in additional roles can be created. That has tended to be the problem in the past. 

“We need to change” is not the same as “and this is how we will do it.” After all, investors have punished connectivity providers who unsuccessfully try to diversify, over and over again. Changing or diversifying always is applauded if it succeeds. But fail and executives will be urged to “stick to what they know.” Been there. Done that.