Buyers of infrastructure and services to use artificial intelligence might be forgiven their angst about payback or monetization of those investments. Sellers have few such qualms.
Roughly the same argument happens around monetization of 5G services: executives complain that they have spent a lot on 5G and have perhaps not seen the financial returns they were expecting, in terms of new or higher revenues.
That is not a new problem, and our experience with fiber-to-home and 5G provides instructive insight.
For some of us, the debate is an old one. In the mid-1990s, for example, it would not have been hard to find an argument about the payback from fiber-to-home networks, either. In the specific context of new competition between telcos and cable operators for voice, internet access and entertainment revenues, the argument was that FTTH would allow telcos to compete with cable in internet access and video, while cable operators took market share in voice.
Then, as now, the issue was the new investments would enable assaults on various markets. Assuming a rough split of new internet access share, telcos expected to take share from cable in video services, while cable two-way networks took some telco voice share.
Financial analysts and operating executives might have hoped for higher returns, but essentially the rationale came down to an existential argument: “do you want to remain in business or not?” Without FTTH upgrades, few, if any, telcos could expect to survive against competitors able to supply hundreds of megabits to gigabits per second home broadband speeds.
That argument applies to 5G investments and clearly will apply to AI investments as well. Though many expect new revenues, use cases, products and services to be possible, the bottom line is that the new investments essentially allow firms to “remain in business.”
“You get to keep your business” might not be highly appealing, in one sense. One would rather be able to claim that investments will produce high financial returns.
But that is not really the choice. The choice is “keep your business or go out of business.” The new investments in 4G and AI are essentially strategic and existential; not fully driven by traditional “return on investment” criteria.
All that noted, some segments of each value chain will have an easier time showing results. As always with a new technology, the initial investments are required to enable use of the technology, and that often means infrastructure suppliers are first to benefit.
If one agrees that the artificial intelligence market can be viewed as consisting of three layers of infrastructure; models and applications, as do analysts at UBS, value creation and supplier revenue also are in layers. As generally is the case for software layers, AI involves layers that also drive or dictate business, revenue and monetization models.
The most-direct monetization will happen at the infrastructure layer, involving direct purchase of hardware, software and capabilities as a service. Nvidia and other creators of graphics processing units and acceleration hardware, as well as servers, are in this category.
Monetization possibilities are mostly direct, in the form of licenses and subscriptions, at the model layer, with some possible indirect monetization for open source models. Subscriptions to use OpenAI; Copilot or Gemini are in this category.
At the applications layer, monetization will mostly be indirect, in the form of improved existing products and services. UBS estimates “enabling” layer products and services including semiconductor production; chip design, cloud and data centers, and companies involved in power supply will generate at least $185 billion in 2027, with total segment revenues closer to $331 billion.
Companies developing large language models and those that own data assets that can be turned into intelligence
Application layer: The companies which embed the tools from the intelligence layer into specific use cases. This layer likely offers the largest monetization potential over time, yet this opportunity is difficult to quantify at this early stage. Presently, the report expects a directly addressable market of USD 395 billion in revenue opportunities for the application layer by 2027.
In the 5G markets, one might note a similar trend. The clearest initial winners were the suppliers of 5G network infrastructure, such as Ericsson and Nokia; construction firms and so forth.
Typically, it takes longer for application success to be discovered.
In that regard, the salient example of direct new 5G revenue is fixed wireless for home broadband. Since about 2022, virtually all net account additions in the U.S. home broadband market have been supplied by fixed wireless platforms.
Other gains attributable to 5G are mostly indirect or hard to quantify, since in most markets supporting 5G services, all the providers offer 5G. In some markets the quantities of various spectrum resources might provide an advantage to one or more providers, such as in the U.S. market, where T-Mobile’s greater trove of mid-band spectrum arguably has allowed it to take market share from the other leading providers.
Still, over time, most of the value of 5G or AI, for most applications, use cases and users, is likely to be realized in more-subtle and indirect ways.
No comments:
Post a Comment