Can Hyperscalers Replace Telcos?

Some predictions have a ring of truth, based in part on history.  One such prediction is that the telecom service provider business could be taken over by hyperscale cloud computing suppliers. 

“I really do believe the hyperscalers are going to become the new telecom providers going forward,” says John Baker, Mavenir SVP.

“Apart from the physical radio that goes on a tower, everything we’re doing now follows the data center model, and these guys know how to manage data centers, software, and applications,” Baker said. “If you look at open RAN essentially as being a collection of applications that run on a server, then it really is falling into their camp.”

One might argue that view is a bit of technological centrism. We sometimes believe the “best technology” always wins." it does not. Sometimes the winner has the better ecosystem. Sometimes the best technology does not offer the highest value. 

Obviously, there are many consumer acceptance issues that shape adoption as well. Brand preferences matter, in that regard. So does customer experience and the complexity of buyer behavior.  

Still, the arenas where hyperscalers and app providers compete has grown over the past couple of decades. The number of roles in the ecosystem where app providers are competing with connectivity providers has grown. Facebook provides its own satellite access services in Africa. 

For example, Google operates one of the world’s biggest collections of undersea networks, sells fiber to the home services, fixed wireless access services, balloon-based internet access and mobile phone service. 

source: GSMA

Facebook, through the Telecom Infra Project, shapes the development of new connectivity platforms and network elements. Though Apple is the biggest supplier of consumer devices, Amazon and Google both have significant roles in tablet or phone markets, or both. And all of those firms are leading providers of smart speaker systems. 

The point is that boundaries between industry segments now are more porous than they used to be. That is the logic behind speculating whether one or more hyperscalers might eventually compete more extensively with connectivity providers. 

While one might disagree with the extent of competition, we can note that competition in any established market generally begins at the revenue edges and gradually works toward the core revenue base. 

MCI initially competed with AT&T only on a few long haul transport routes, but gradually became a full rival to AT&T in every product segment related to long distance services: wholesale and retail; voice and data. 

Skype originally only was a substitute for community PC-to-PC calling. It gradually became a full substitute for calling between phones. Streaming services originally targeted only non-real-time pre-recorded content. Increasingly, some streaming services are adding live sports and other real-time content. 

Streaming services once offered a menu of movies and older TV series. Now streaming services create original programming, offer first-run movies and increasingly could become a supplier of first-run release content as well. 

Cable TV began life as an importer of distant TV signals. It eventually became a full-fledged supplier of new networks and original programming far beyond traditional broadcast TV. 

Google remains a relatively niche supplier of internet access and mobile phone service. But that is consistent with the traditional attack trajectory, is it not?

PTC'21 Will be Quite Different This Year

Tariff and Price Complexity is a Strategy

There is a reason why consumer internet access or mobile service prices are intentionally complex: complexity makes it difficult to compare, across providers. Complexity often is facilitated by unbundling service into components. 

Unbundled services--by definition--make price comparisons more difficult. But unbundling allows many different value combinations and feature bundles to be created, deemphasizing like-to-like value comparisons.  

That is not to deny consumer welfare gains in the form of cost savings from bundle prices. But service providers arguably do not sell bundles primarily because it benefits consumers. Rather, bundles benefit the suppliers as well.  

In other words, the complexity of consumer communications service billing is intentional, and designed to optimize revenue for service providers. Complexity does so by limiting the ease of comparison shopping by price. 

Consider multi-product bundles, which literally obscure what consumers actually are paying, from the consumers themselves. By definition, a three-product bundle, sold for one price, not only offers a discount over separate retail prices, but also obscures what each component actually costs. 

Sure, one can try and approximate the savings by comparing the total cost of the bundle price against the separate, a la carte prices of the constituent bundle components, and then apply a flat-rate discount against each separate service. 

That probably is not how the service providers see matters. Instead, the bundles are likely designed to optimize overall revenue, with key benefits coming from reduced churn or account longevity. 

If it produces higher overall revenue and higher profit margins, one service might be heavily merchandised to protect higher take rates for another service that produces more revenue or higher profit margin. 

To be sure, component value is changing. Internet access is more important; linear video is less important. Most consumers prefer to use their mobile phones for voice, rather than landline voice service. Streaming is the growing choice for video service, reducing the value and revenue from linear video subscriptions. 

Still, unbundling value also creates many different ways to rebundle value that obscure price comparisons and create value incentives for purchase of more-expensive plans. Mobile internet plans differ on the amount and use of tethering, for example. The cheapest plans might not allow tethering, or restrict its usage. 

“Unlimited data usage” plans tend to be the most expensive offered by a mobile operator, for example. Some service providers require purchase of an unlimited plan to use 5G. 

The point is that complexity and unbundling also create optimized revenue scenarios for service providers that obscure price comparisons. 

Hard to Say What a Typical U.S. Internet Access Service Costs

With the caveat that the results are not generated by a statistically randomized poll, a survey conducted by cabletv.com suggests the difficulty of figuring out what consumers actually pay for internet access. 

The reason is bundles, which effectively obscure what consumers actually are paying, from the consumers themselves. So if asked, consumers buying bundles including internet access have to guess at what they are paying. 

For starters, cable operators have about 70 percent of the installed base of fixed network internet access accounts, so cable customer behavior is disproportionately important. Among cable operators, bundles including internet access range from a high of 73 percent to a low of 45 percent. 

source: cabletv.com. 

Effectively, that means nearly three quarters to half of cable internet access customers do not really know what they pay. So the self report data is, by definition, suspect. 

Third party efforts to estimate those payments generally have to rely on higher published stand-alone retail prices, which most cable customers are not paying. On the other hand, some estimates include promotional plan pricing, which also can skew the results. 

All that noted, the “average” U.S. price for internet access is about $50 a month. Using various methods of comparing global prices, though, U.S. prices seem to be right at the global average, paradoxical though that may seem. “Average” is always a statistical matter for anything related to the internet, and that holds true for consumer access prices as well. 

One has to adjust for currency effects, different costs of living  in various countries, choose which plans to compare, adjust for promotions and discounts as well as speed differentials, for example. 

If “average”  purchased speeds are 150 Mbps in one country, but 10 Mbps in another, what does the “same effective price” mean? Is price per megabit-per-second a better metric? And how does one average asymmetrical buying patterns where significant percentages of customers buy service at speeds above 300 Mbps and also at 25 Mbps?

Using the purchasing power parity method of normalizing prices across countries, U.S. fixed network internet access is cheaper than the global average of $73 a month. So average U.S. prices are significantly lower than the global average. 

Even as a "Platform," Telcos Would Not Escape Near Zero Pricing

The reality of very low and declining per-unit prices is well attested in the connectivity business. Many suggest a way out of the conundrum is for at least some connectivity providers to transform themselves as platforms. 

Ignore for the moment whether this is generally possible, and to what extent. 

Life as a platform would ultimately be based on very low per-unit prices. In fact, as many platforms feature zero marginal cost, they also tend towards near zero pricing. 

Virtually all platforms feature lower prices per unit than rival pipe businesses, for a number of reasons. Typically making extensive use of internet and computing resources to radically lower transaction and information discovery costs, etailing platforms inevitably push cost out of retail transactions. Platforms reduce friction. 

In other cases, platforms are able to mobilize and put into commercial use assets that otherwise lie fallow. Uber provides a good example. Personally-owned vehicles tend to sit parked and unused 95 percent of the time. Uber allows those otherwise idle assets to be put to commercial use. 

And though firms often are urged to become platforms, few actually can do so, and not for reasons of technology deployment, skill or type of product. Successful platforms are relatively rare because they require scale, and few businesses can afford to invest to scale. 

Most firms in the connectivity business will not be able to transform as platforms, leaving only other possible options. If one believes that prices for telecom products are destined to keep declining, or that more for the same price is the trend, then there are a couple of logical ways to “solve” such problems. 

Firms might try to gain scale to lower unit costs, change the cost model in other ways to enhance profitability, exit the business or change the game being played. Moving “up the stack,” across the ecosystem or into new or adjacent roles within the value chain can “change the game.” That is the strategy behind Comcast and AT&T moving into the content ownership business, or moves by other tier-one service providers into new lines of business outside the connectivity core. 

That is one way to attempt to escape the trap of marginal cost pricing, which might be the connectivity industry’s existential problem. 

But it also is reasonable to assume that even a successful shift to a platform model will be based on near zero marginal cost, and near zero pricing. The reason is simply that most platforms also feature near zero pricing.

U.S. Broadband Speeds have Improved Faster than Expected

It often is useful to revisit bandwidth or cost predictions periodically, to test the predictive value of algorithms. Back in 2013, for example, Technology Futures, using several standard industry algorithms, predicted that, by 2020, half of U.S. consumers would be buying 100 Mbps internet access connections.

source: Technology Futures

Those same tools suggested that about 10 percent of consumers would be buying 50 Mbps connections, while nearly 24 percent would be buying 24 Mbps service. 

In 2020, according to Openvault, about 62 percent of U.S. consumers were buying internet access at speeds of at least 100 Mbps. 24 percent were purchasing service at speeds of at least 50 Mbps, while 12 percent were buying services running between 20 Mbps and 40 Mbps. 

source: Openvault

So actual behavior exceeded the 2023 predictions made by Technology Futures, using the Fisher-Pry model for technology adoption and Gompertz growth forecasting tools. 

There is no particular reason to believe gigabit speed adoption will not follow the Fisher-Pry suggested growth path, either. It should develop as an “S curve,” as did all prior generations (1.5 Mbps, 6 Mbps, 24 Mbps, 50 Mbps, 100 Mbps, for example). 

Some 154.2 million U.S. residents can buy gigabit internet access service from cable operators,  according to Viavi data. If the U.S. population is 328.2 million, that implies 47 percent of people can buy the product. 

The NCTA claims 80 percent of U.S. households can buy gigabit service from cable operators, however. 

Where it comes to predicting typical U.S. consumer internet access speeds, the industry-standard models might have been too conservative.

Pipe or Platform?

For all the talk of the value of “platforms” in the internet ecosystem, platforms are much harder to create than the typical “pipes” business. So the “good advice” to “become a platform” might apply to only a smallish percentage of all connectivity provider businesses. 

A platform business model essentially involves becoming an exchange or marketplace. A pipe model requires a firm to be a direct supplier of some essential input in the value chain.

A platform functions as a matchmaker, bringing buyers and sellers together, but classically not owning the products sold on the exchange. A pipe business creates and then sells a product directly to customers. Amazon is a platform; telcos and infrastructure suppliers are pipes. 

Platform creation is not especially easy for a connectivity services provider. If you think about every business as either a “pipe” or a “platform,” then most businesses are “pipes.” They create a specific set of products and sell them to customers. That is a classic “one-sided market.”

It arguably is easier to build and sustain a business using a “pipe” model than a “platform” model for a number of reasons, but primary among those reasons is that a pipe model only requires knowing what problems the potential buyer faces, and then coming up with a workable and affordable solution. 

Creating a platform model requires understanding the problems that at least two sets of participants may have, and then devising a way to solve each participant’s problem. At a crude level, that means a platform’s business value challenge is twice as hard as that facing a pipe provider. 

Of course, for all the attractiveness of a platform model, it can be argued that creating a platform is much more capital and scale intensive than operating an alternative pipe model. In fact, small firms have almost no choice but to operate using a pipes model. 

Also, some businesses we think of as “platforms” might actually operate as pipes. Amazon’s e-commerce operations are a classic platform. But Amazon Web Services arguably operates as a pipe: it sells customers computing or storage services that actually are “pipe” products.

In other words, AWS is an example of one supplier (AWS) selling a specific product directly to one customer. One might say the same for a single retail brand selling online. That is a pipe model, not a platform, as it features one seller, making transactions with a single buyer. 

Another way of stating this is that a pipe model has customers interacting with a seller; a platform has users interacting with each other. 

Also, the value for different sets of participants might be asymmetrical, which is why platform providers sometimes have to create incentives for one side of the platform to participate. Uber pays its drivers, for example, to ensure adequate availability for riders. AirBNB compensates rental partners for their participation. Expedia pays booking commissions to its travel providers. 

Yet another way of illustrating the difference is that a platform orchestrates interactions and value. In fact, a platform’s value may derive in large part from the actions and features provided by a host of ecosystem participants. Facebook’s content is created by user members. Amazon’s customer reviews are a source of value for e-tailing buyers. 

Consumers and producers can swap roles on a platform. Users can ride with Uber today and drive for it tomorrow; travelers can stay with AirBNB one night and serve as hosts for other customers the next. Customers of pipe businesses--an airline, router or phone suppliers, grocery stores-- cannot do so

You might be aware that firms in an ecosystem often move into adjacent roles within the ecosystem, or challenge incumbents in other ecosystems. Though that can be done by either a pipe or a platform, it might be easier for a platform. 

Network effects and communities both have business value, and platforms might be able to more easily transfer scale advantages to new lines of business. Pipe-oriented firms can do this as well. The difference might be the ease of scaling platform assets, which by definition require wide participation by lots of different firms and industry segments.

A pipe-oriented business will have advantages in its current customer base, ex-customers and suppliers. A platform might already have relationships with supply chain, payment, logistics, marketing or user relationships that are far broader. 

In a pipe business, the seller “creates value” represented by the product sold. In a platform business, value can be created in different ways, such as recommendations. In that case, neither the platform itself or the sellers of products are creating the value, but the customers or users. 

The other issue is that any business using the internet is itself participating on a platform, though perhaps indirectly. And as is true for most products sold or exchanged over the internet, there is a tendency for excess profit to be removed, which means any product available on the internet--by pipe or platform--tends to move towards zero margin. 

Also, many businesses may operate partly as a pipe and partly as a platform. Perhaps the Apple iPhone provides a good example of that.