Does ChatGPT Pose New Legal Risks for Search Engines?

Ignoring for the moment possible regulatory moves on the order of the breakup of the monopoly Bell system (AT&T divestiture) or the Microsoft consent decree, ChatGPT itself poses some new possible issues for Google and other firms that incorporate ChatGPT as part of their operations. 

Though a subtle matter, ChatGPT arguably changes ecosystem roles. Today, Google’s search responses arguably remain in the realm of “finding products and content” rather than “creating products and content.”

A search query produces results created by third parties. A ChatGPT query produces something very different: a specific answer and formulation. The analogy might be that ChatGPT makes its sponsor a content creator; a publisher. 

Search, in principle, only makes the provider an aggregator of third party content. 

In some countries, that distinction can be significant. Today, in some countries, aggregators are not legally responsible for content third parties create. But publishers of content (newspapers, radio and TV stations, magazines, news and opinion sites) can legally be held responsible for what they choose to publish. 

In some instances, that is anything but a subtle difference. The legal defense for an aggregator is that laws protect it from liability when others say things. Of course, in some countries publishers also have protection from prosecution when “reporting the news” or “offering an opinion,” with the exception of willful publishing of material known to be untrue, and injurious to people who are not public figures. 

So ChatGPT raises new issues about whether aggregators and search engines are “merely” conduits for third party information or now become something more like “speakers” in their own right. 

Whether this change ultimately changes legal liabilities is unclear at the moment. Whether those possible levels of legal and financial exposure affect content moderation policies likewise is unclear. 

But it is a new area of uncertainty.

Changes in U.S. FTTH Demand and Supply Sides

There are several reasons--both supply side and demand side--why U.S. “fiber to home” business models appear to have changed. 

Perhaps oddly, fundamental demand for home broadband, though higher than ever, also provides less of the revenue to build the networks.

As important as the fiber-to-home business is, it is responsible for less than 10 percent of AT&T revenues. In the fourth quarter of 2022, for example, mobility drove nearly 69 percent of total revenue. 

In the fourth quarter of 2022. AT&T earned $31.3 billion. Mobility generated $21.5 billion of that amount. The fixed networks business generated $8.8 billion. Consumer fixed network services generated $3.3 billion or so. 

Of course, not all contestants are similarly situated. For many competitive internet service providers, revenue does largely depend almost exclusively on home broadband. Again oddly, revenue potential for such ISPs also seems to have declined. 

Where designers once assumed FTTH per-customer home revenue in triple digits ($130, for example), they now assume revenue in the $50 to $70 a month range. That might seem to eviscerate the business case, if network costs are in the $800 range with additional costs to connect actual customers in the $600 to $725 range, with take rates ranging from 20 percent up to about 40 percent. 

Only a firm with low overhead can make money sustainably at 20 percent adoption rates. For larger firms, adoption in the 40-percent range is likely required. At a high level, AT&T has been saying FTTH payback models work at $50 a month ARPU and penetration of 50 percent, though revenue from targeted newbuilds now exceed those figures, AT&T says. 

It is one thing for a smaller ISP to contemplate building an FTTH network and sustaining itself solely on such revenues. It is quite another matter for a dominant firm in a local area (Comcast, Charter, AT&T, Verizon, Lumen, Frontier, Brightspeed). 

Strategic concerns also matter, however. Even if the fixed networks business generates 10 percent of total revenue, that revenue still matters. Without the FTTH upgrade, AT&T risks losing that revenue and profit margin and cash flow contribution. 

In other words, even if never stated so starkly, unless the FTTH upgrade is made, AT&T and others risk losing their fixed networks business to competitors. 

At the same time, though harder to quantify, the payback model for deep-fiber networks can come in other ways. If small cell mobile networks require deep fiber networks, then business value comes also from the value of the backhaul network. So “fiber to the tower” and “fiber to the radio site” become elements of the payback model. 

Fiber access networks also support the business customer revenue stream. For AT&T, fourth quarter 2022 fixed networks business revenue was $5.6 billion, or about 18 percent of total revenue. So “fiber to the business” arguably drives almost twice the revenue as home broadband does, for AT&T. 

In other words, the same network supporting home broadband also contributes to support of the mobility business and business customer revenue streams. 

All that makes for a more-complicated payback analysis for any sizable contestant with dominant mobile revenues. Though the home broadband payback has to be there, the value of what we used to call “FTTH” has to be justified in other ways. 

Smaller ISPs might be able to justify an FTTH network on the basis of home broadband services alone, with 20 percent take rates. It is not so clear a large dominant service provider can hope to do so unless it can reach 40 percent or higher take rates, assuming revenue per account in the $50 to $70 range. 

And even when it does so, total deep fiber network value can hinge on other value contributions. 

Still, there are additional considerations. Supply side support from the federal government can reduce the cost of rural networks builds by 20 percent to 30 percent, which aids the payback model. 

Joint ventures of various types provide similar benefits, at the cost of possibly further reducing net revenue upside. 

And though it is an indirect input, many private equity firms are willing to invest in deep fiber projects with a rather simple formula: buy assets at a five times to six times revenue multiple and upgrade with FTTH to produce an asset selling at 10 times to 11 times revenue multiples. 

Demand side drivers also have changed a bit as well, beyond the “need” for internet access. 

The Affordable Connectivity Program provides a $30 a month subsidy for low-income buyers. That subsidy can be used to buy basic or faster services, and increases demand for internet access. 

In some cases, that means new FTTH facilities benefit both from 20 percent to 30 percent lower build costs, plus $30 a month in consumption subsidies for lower-income households. All those are new elements in payback models that improve the business case on both demand and supply sides.

Computing Now is Inseparable from Communications

Sometimes we only understand the value of something when it suddenly becomes unavailable. 

What is the value of one’s mobile phone, personal computer, applications, television, smartwatch, sensors or transaction appliances when power and internet connectivity are lost? If one could quantify value, it would be a number close to zero, in the absence of power and internet connectivity. 

It is estimated that intentional shutdowns of the internet by governments alone caused $24 billion in economic losses in 2022, for example. Deloitte researchers estimated in 2016 that a one-day internet outage affecting 10 million people in a country with high gross domestic product costs $24 million. 

source: Deloitte 

The point is that modern computing is inseparable from internet connectivity. That applies equally to end users as well as application and service providers; data centers and content owners; transaction platforms, logistics, transport, media and retail operations.

Most computing in the mainframe and minicomputer eras, followed by the early personal computer era, did not rely centrally on wide area networking, though local area connections often were important or essential. That largely remained true in the client-server era. But all that changed when computing became highly distributed in the internet, web, cloud computing and mobile computing periods, when resources were, by definition, remote. 

Paul M. Veillard 

As all media types became digital, content consumption and creation also were decentralized. The obvious implication is that modern computing cannot be divorced from networks and communication. 

Though much computing and processing happens at the edge, in an organization's private data center, onboard a PC, sensor or mobile device, most workloads now require remote processing. Hence the need for good connectivity, ranging from home broadband to mobile networks to metro capacity, inter-city links and subsea networks. 

source: Telegeography 

Why it is Hard to Move up the Stack, Much Easier to Move Down the Stack

Vertical integration and adding new roles in any value chain are traditional ways firms seek to increase value or control costs and value. As much as connectivity providers talk about “moving up the stack,” app providers also can move “down the stack.”

And we might as well just acknowledge that it is easier for an app provider to move down the stack than for a connectivity provider to move up the stack. 

The reasons are somewhat obvious if you think about the issue long enough. Any app, service or product provider already knows lots about their customers. In other words, a business operating above the app level knows what its customers want, why they buy and how much they prefer to pay. 

A connectivity provider has to learn what its connectivity customers want, but typically has no direct knowledge of the intimate details of how those connectivity customers actually create value in their businesses.

In other words, firms operating at higher levels of the stack are intimately involved with the actual business functions connectivity supports. Transport and computing functions at the lower levels are less involved--if involved at all--in the higher level business processes. 

source: Vermont IT Group 

Bluntly, a connectivity supplier only knows what a class or type of customer typically wants to buy, in terms of computing and connectivity services, but has no direct and detailed knowledge of the connectivity customer’s actual business. 

That is why connectivity provider enterprise sales forces have to build domain knowledge. Those in the domain already know all that, in detail. 

It no longer is unusual for an app provider to become an access provider, for example. Google Fiber provides only one example, operating as a retail internet service provider and as an owner and builder of substantial wide area networks across the globe. Meta, Microsoft and other app providers also are anchor tenants if not full owners of WAN assets. 

Tucows, originally a domain name registrar, has become both a mobile services provider and now a gigabit Internet service provider. 

There also is movement by new providers into existing connectivity roles. Cable companies, satellite companies and other original equipment manufacturers likewise have moved into additional parts of the retail connectivity services business, ranging from internet access to mobile services. 

Facebook now leases transponder time to support internet access operations in sub-Saharan Africa and sponsors the Telecom Infra Project that develops new open source tools across the connectivity ecosystem. 

In other words, it is far easier to move down the stack than to move up the stack. 

Where are the Enduring Business Moats for Fixed, Mobile Operators?

As more digital infrastructure assets shift into private equity or institutional investor hands, a logical question is whether there is a change in the perceived business value of network and facilities ownership, or simply a change in network costs and business models. 

In the past, ownership of scarce network access assets was considered to provide value because the cost of such facilities was so high that a business moat was created for the owner of the assets. 

Of course, 50 years ago such business moats also were complemented by government  monopolies as well. These days, under competitive frameworks, only the business moat matters. 

With the caveat that business strategies can change over time, there has been some movement towards layered, disaggregated facilities models in both fixed and mobile realms. In the mobility business, open radio access networks, virtualized functions, tower asset sales and the whole mobile virtual network operator model provide examples of a shift away from vertically-integrated formats.

In the fixed networks business, some countries have moved to “single wholesale network” approaches, which can, in some cases, eliminate the business value of facilities ownership nearly entirely, as all retailers use the same network, with the same features and underlying network cost structures. 

In markets where a range of options are possible, more instances of joint ventures are visible, in large part a means of reducing the risk and cost of new fiber-to-location facilities while arguably also speeding deployment timetables. 

In such cases, it is not yet possible to say the perceived value of an access network (scarcity value) is diminished. In fact, the need to deploy first or early, before other contestants can deploy their own facilities, might tacitly support the notion that the network remains a scarce asset with moat value. 

In other words, if deploying the first FTTH network leads to 40 percent take rates, while following networks only manage to gain about 20 percent each, then the scarcity value remains, even if not as robustly as in the monopoly era. 

That noted, in some cases, fixed network operators have traded a local monopoly in exchange for government permission to expand on a wider geographic basis. Few remember it now, but Rochester Telephone in 1993 opened its network to competitors in exchange for the right to enter the long distance business. 

Singtel essentially did the same thing later, giving up its Singapore monopoly for freedom to expand elsewhere in South and Southeast Asia. 

Those are examples of strategic decisions that do not directly bear on the issue of the business value of access networks produced by their scarcity. 

At least so far, mobile operators have been able to shed tower assets without harming their franchises. They should be able to virtualize their networks--assuming no business or technology missteps--without marketplace damage. 

It is not so clear yet how value could change in the fixed networks business. So far, most joint ventures to build FTTH infrastructure arguably are driven by a need to build faster than the competition, so the value of the joint venture is lower cost and time to market, and not a change in the perception of value. 

In the data center market, changes in ownership have not necessarily signaled a change in belief about scarcity value, either. Whether an owner is one commercial entity or another, or even if ownership changes from private to public, or public back to private, does not intrinsically change the scarcity value of the assets. 

Indeed, the surrounding issues of water scarcity; energy consumption; location and scale all suggest scarcity value remains key. First movers in any geography often have advantages. Environmental and social constraints are starting to limit unbridled expansion as well. 

Still, in the data center market, facilities ownership does not seem to have lost any luster, with the obvious caveat that even the largest hyperscale data center operators lease facilities all the time, rather than building and owning their own assets. 

That pattern appears more open to change in the fixed networks and mobile businesses, where facilities scarcity value has essentially been eliminated, in some markets where a wholesale-only model prevails and a neutral third party provides the network facilities. 

In other markets where a former incumbent also owns the wholesale facilities, some additional scarcity value is retained by the former incumbent. In markets with no restrictions on facilities-based competition, scarcity value remains, but perhaps at a lower level than in the past. 

A typical pattern is ownership in core geographies and sharing or leasing in out-of-region markets. Again, that speaks more to conservation of capital than scarcity value. 

Still, questions about scarcity value seem destined to grow as formerly-vertically-integrated firms adapt to the possibility of a layered approach to their operations. The debates access providers now have about outsourcing operations support to public cloud services providers is an example of that deepening debate. 

At its heart, all such debates are about value, and where value lies.

Service "to Persons" has Key Revenue Implications, Compared to Service "to a Location"

For anyone who has covered or analyzed the U.S. connectivity business for some decades, it now is somewhat shocking how much revenue is produced by the mobility business, compared to the fixed networks business, even granting the importance of the fixed network for business and home broadband. 

Keep in mind one salient element of each business: fixed services are “to a location” while mobile services are “to a person” or “to a sensor.” So revenue per locations is one number while mobile revenue is a compound number based on the humans and network-connected sensors at a given location, 

In other words, mobile revenue can easily be 2.5 to 5 times the revenue of a fixed network connection. 

AT&T mobility average revenue per postpaid account was $55.43 in the fourth quarter of 2023, while home broadband accounts served by optical fiber had $64.82 average revenue per account. 

So you can see the total revenue per account implications when multiple mobile accounts are purchased, compared to a single home broadband connection. 

AT&T, for example, earned $31.3 billion in the fourth quarter of 2022. Mobility generated $21.5 billion of that amount. The fixed networks business generated $8.8 billion. In other words, mobility drove nearly 69 percent of total revenue. 

source: AT&T

The fixed networks business revenue was $5.6 billion, or about 18 percent of revenue, while consumer fixed network revenues represented about 10 percent of total revenue. As important as the fiber-to-home business is, it is responsible for less than 10 percent of AT&T revenues, as voice revenues and copper access are part of those revenues. 

 

 

source: AT&T

Of course, each product has different profit margins, so revenue does not tell the whole story. Business service, consumer broadband (fiber versus copper) and consumer and business mobility all likely have distinct profit margins. 

So smaller revenue contributions might generate higher amounts of firm profit. Still, the law of large numbers is evident. A one-percent improvement in mobility segment revenues or profit should have higher firm impact than a similar one-percent increase in either business or consumer fixed network services. 

Home broadband probably has the highest margins, and those margins might be getting better. Because of federal government subsidies for fiber-to-home construction. Those subsidies might reduce the cost of FTTH builds by 20 percent, in rural and other difficult-to-serve areas. 

We Used to be Able to Count Trans-Atlantic bandwidth in T1s

Back in 1979 when the PTC was formed, trans-Atantic bandwidth was about 1,000 Mbps total, or about 647 T1 circuits. In early 2023 trans-Atlantic capacity is likely above 75 Tbps. 

source: AI Impacts