Can AI Follow the Models of Search and E-Commerce to Create New Products or Monetization Models?

Some idea of the disruption AI might eventually support can be seen in business model possibilities in the computing markets for hardware and software. 

Traditionally, software and hardware firms made their revenue selling those products to customers who wanted software and hardware solutions for their business or personal uses. 

It would at one time have been odd to suggest that a major technology products company would make its money not from selling software or hardware in a direct sense, but instead supporting applied technology with a revenue model based on advertising or e-commerce, for example.

So Amazon, Uber, Airbnb and Doordash, which might be viewed as “technology” firms, make their money by retailing transactions. Alphabet mostly makes its money from advertising, as does Meta. 

And even many traditional hardware and software firms now make most of their revenues from subscriptions rather than outright sale of software or hardware products. In a sense, many former “hardware and software” firms now are “service providers” akin to providers of communications, fresh water, wastewater, electricity, natural gas or trash removal services. 

But that shift to “services” also has allowed new products to emerge, notably software as a service; computing cycles as a service; data storage as a service and so forth. Uber represents “transportation as a service” akin to passenger use of railroads, airlines, cruise lines, ferries or taxicabs. 

And perhaps that offers clues about where to look for the emergence of distinctively-new products AI will enable. Ignore for the moment AI as an infrastructure business for ChatGPT or other large language model suppliers. 

In some cases  the actual technology products people and businesses use will emerge. In other cases it might be the monetization of value that changes. Two decades ago, for example, firms experimented with the idea of connectivity services supported by advertising. 

More recently, Amazon experimented with the idea of ad-supported phone service. Free Mobile, the French mobile operator, has offered a free mobile plan with ads.

FreedomPop, the U.S. mobile operator that was acquired by Red Pocket Mobile, offers a free mobile plan with ads. The company generates revenue from advertising displayed on the phones of its users, as well as from premium services such as additional data and international calling.

AdMobilizer is a company that provides a platform for mobile operators to offer ad-supported mobile plans. The company works with mobile operators in over 50 countries to offer ad-supported plans to their users.

Such efforts have not been huge successes, but the concept is there. If software services can be supported by advertising or commerce, why not connectivity services? 

The precedents are there: technology products supported by advertising or e-commerce, rather than by direct sales of technology (hardware, appliances, software licenses). 

Company	Revenue contribution from activities other than direct sales of hardware or software products	Percentage of total revenue from products other than direct sales of hardware or software as a service
Amazon	E-commerce, advertising, AWS	>90%
Alphabet (Google)	Advertising	86%
Microsoft	Azure, Office 365	70%
Meta	Advertising	97%
Salesforce	SaaS	97%
Shopify	SaaS	98%
Stripe	Payment processing	99%
Twilio	Cloud communications	97%
Uber	Transportation	100%
Airbnb	Accommodation	100%
DoorDash	Food delivery	100%

Since advertising opportunities tend to require volume ("eyeballs" or attention), it seems likely that new types of AI products supported by advertising will lie in the consumer markets. 

How Much AI Return from "Efficiency," Compared to "More Revenue?"

Investors and company executives rightly must think about how increased investments in artificial intelligence will produce a return on those investments. In most cases, the current thinking is that applied AI will allow firms to reduce operating costs or boost customer experience. 

That is likely to be the most-immediate form of payback in most cases. 

But there is some thinking that applied AI can boost revenues in some way related to higher rates of customer acquisition, possibly lower churn or higher revenue per account. 

For example, AI is expected to support:

• Personalized marketing and recommendations that boost sales

• Product development based on customer insights

• Pricing optimization based on customer buying behavior

• Fraud detection and prevention

• Operational efficiency

Amazon uses AI to personalize its product recommendations and to develop new products, while Netflix uses AI to recommend movies and TV shows to its users.

Spotify and Airbnb do the same for making music recommendations or matching lodging vendors with guests. 

Walmart uses AI to optimize its inventory levels and to improve its supply chain. All those are essentially process improvements. 

Any firm can use AI to analyze product reviews, allowing it to identify product weaknesses or strengths. 

Thinking is less developed on ways AI can be used to create entirely new products. It is one matter to investigate ways AI can make existing products better, boost profit margins or cut costs when creating, distributing, marketing or selling those products. 

The analogy might be the creation of “search” or “social media,” which really were entirely new products. Those were big new categories of products. In most cases, “direct to consumer” e-commerce was not entirely new, but simply a process improvement, even if it radically reshaped distribution mechanisms. 

That is not to deny that such DTC changes upend industry business models. Look at the changes in the video, audio and print media industries over the past few decades, where streaming is displacing linear video; where streaming has displaced physical media; where online delivery has replaced physical content media. 

Historically, most new technologies allowed us to reshape business processes to make them “more efficient” or “automated” or “self provisioned” or “virtual.” And while that is likely to be the case for applied AI, the big new industries and firms will move beyond that, as did search and social media and possibly cloud computing. 

The UGC Precedent for AI Disruption

Big technology disruptions are typically accompanied by predictions of big business model or revenue disruption. And disruption often does occur. But the degree of disruption varies. Rarely does an internet substitute completely eradicate a legacy model, but rarely does an internet substitute fail to significantly affect the legacy business. 

That is very likely to be the case for applied artificial intelligence as well. AI will augment in most cases. Rarely will AI completely replace a legacy solution, at least for the coming decade or so. But that could change dramatically after a decade of deployment. 

Consider predictions for changes in content production after user-generated content was widely possible. A few believe professional content production would not be affected, but more observers predicted widespread substitution.

In practice, UGC has underpinned the success of social media. But it has not replaced professional content production. Generally speaking, UGC has increased the supply of content and allowed the building of revenue models on UGC, without displacing professional content.

The caveat is that professional content now often uses social media as a distribution channel. 

Prediction	Predictor	Date of Prediction	Actual Result
90% of all content will be UGC by 2020.	Chris Anderson, author of "The Long Tail"	2006	UGC now accounts for about 25% of all content.
50% of all news will be UGC by 2020.	Jeff Jarvis, author of "What Would Google Do?"	2009	UGC now accounts for about 20% of all news.
25% of all TV shows will be UGC by 2020.	Clay Shirky, author of "Cognitive Surplus"	2010	UGC now accounts for about 10% of all TV shows.
UGC will displace 70% of professionally-created content by 2025.	Andrew Keen	2010
UGC will never fully displace professionally-created content.	Nicholas Carr	2011

Look at the changes in distribution channels wrought by internet mechanisms, for example. 

Video streaming has been an effective substitute for most traditional video subscriptions, for most consumers.  In 2022, 73 percent of U.S. adults subscribed to a streaming service, compared to 65 percent who subscribed to a traditional cable or satellite TV service. 

In 2022, 74 percent of U.S. adults booked their airline tickets online, compared to 17 percent who used a travel agent.

In 2022, 73 percent of US adults booked their hotel rooms online, compared to 13 percent who used a travel agent.

In 2022, U.S. e-commerce sales accounted for 14 percent of total retail sales, up from five percent in 2010.

In 2022, U.S. digital advertising spending accounted for 57 percent of total advertising spending, up from 31 percent in 2010.

Industry	Percentage change in distribution channels from traditional to internet channels, 2012-2022
Video streaming	50 to 60%
Airline ticket sales	60%
Hotel bookings	50%
Retailing in general	30%
Advertising	20%
Marketing	15%

Just as UGC augmented content production (social media is mostly UGC) and rearranged distribution channels (streaming in addition to linear video), so AI is very likely to augment and transform legacy processes rather than replace them.

That arguably also applies to jobs. Demand for some jobs will dip; other new jobs will be created. Most will be affected. The same thing happened when the personal computer revolution happened; when the internet happened. 

But lots of other technology transformations, such as the use of mobile phones or remote work, will mostly modify or reshape behavior and processes, even if some direct creation and elimination happens. 

We should expect artificial intelligence outcomes something along the lines of user-generated content: displacement in some cases; augmentation in some cases; transformation in possibly most cases.

LLM-Assisted Search Would Dramatically Boost GPU Usage and Energy Consumption

Concern about business and revenue models for large language models and generative AI are logical enough, for a variety of reasons. Model creation and inference generation cost money (graphics processor units, GPU as a service, code generation and software engineer support) and energy consumption. 

SemiAnalysis, for example, has estimated that implementing AI similar to ChatGPT in each Google search would require 512,821 of NVIDIA’s A100 HGX servers, totaling 4,102,568 GPUs.

By some estimates, GenAI applied only to the search function for Google, if nearly ubiquitous, would boost costs of such search dramatically. 

source: Semianalysis 

At a power demand of 6.5 kW per server, this would translate into a daily electricity consumption of 80 GWh and an annual consumption of 29.2 TWh, an estimate similar to that made by New Street Research.

New Street Research has estimated that Google would need approximately 400,000 servers to handle search queries, each using an LLM model, which would lead to a daily energy consumption of 62.4 GWh and an annual consumption of 22.8 TWh.

With Google currently processing up to nine billion searches daily, these scenarios would average to an energy consumption of 6.9–8.9 Wh per request. 

source: Joule 

So Far, Nobody Can Imagine the Entirely-New Products, Use Cases, Industries and Revenues AI Will Create

Forecasting technology is perilous, many would tell you, and they’d be right. Keep that in mind as “everyone” starts to opine about the future of AI and what it means for life, government, the economy, industries, products and processes, innovation, education or culture. 

Some trends futurists predicted in 1995 for the internet have not come to pass, and some that seem to be taking decades to occur. As always, we have been wrong about lots of things. 

Prediction	Person Who Made Prediction
The internet will be a fad.	Bill Gates, co-founder of Microsoft
The internet will kill television.	Nicholas Negroponte, architect of the MIT Media Lab
The internet will lead to the end of privacy.	Vinton Cerf, one of the "fathers of the internet"
The internet will be used primarily for education and research.	Clifford Stoll, astronomer and author
The internet will make everyone smarter.	Kevin Kelly, co-founder of Wired magazine

Other general shifts are happening, but inconsistently and over a longer time span than some might have predicted. The internet was believed to revolutionize education, but so far the impact has been more evolutionary. 

.

The internet was believed to make it possible for people to work from anywhere in the world, and that has happened, to an extent. But not every job function can be virtualized, and not every job function has better outcomes for customers or users when virtualized. 

As always, some believed the internet will lead to a more decentralized and egalitarian world. It is not clear how much that has happened, and in what areas of life it has happened. People undeniably have more opportunities to contribute to what we once called “media,” as in “social media.” 

But it is questionable whether that degree of participation or enablement has really led to the world becoming more decentralized or egalitarian. 

For example, forecasters predicted in the mid-1990s that the internet would lead to a more decentralized world, with less reliance on large institutions or firms. However, in reality, the internet arguably has led to a more centralized world, with a few large companies dominating the online landscape in almost any field.

Internet or not, that tends to be the pattern in every industry, globally, in any case, so we were likely naive to believe the laws of economics would change. 

Many predicted that traditional media, entertainment and content industries would disappear, or be replaced by new challengers. Though pressures clearly exist, with product formats shifting from physical to virtual, existing industry leaders have managed to cope, if with difficulty. 

Perhaps the better description is that the internet slowly is forcing a shift of media, content and entertainment business models from physical to virtual, but that the shift is taking decades to happen. Two and a half decades after the early internet became a mass market reality, we still are transitioning from physical to virtual in the value chain. 

But humans seem always to have been poor forecasters. Consider the many examples of smart, informed observers making completely incorrect predictions. 

* "There is practically no chance (that) communications space satellites will be used to provide better telephone, telegraph, television or radio service inside the United States."—T.A.M. Craven, Federal Communications Commission commissioner (1961)

* "I think there is a world market for maybe five computers."—Thomas Watson, chairman of IBM, 1943

* "The Americans have need of the telephone, but we do not. We have plenty of messenger boys."—Sir William Preece, chief engineer, British Post Office, 1876

* "This 'telephone' has too many shortcomings to be seriously considered as a means of communication."—Western Union internal memo, 1876.

* "Television won't be able to hold on to any market it captures after the first six months. People will soon get tired of staring at a plywood box every night."—Darryl Zanuck, 20th Century Fox, 1946

* "Everyone's always asking me when Apple will come out with a cell phone. My answer is, 'Probably never.'"—David Pogue, The New York Times, 2006

* World Economic Forum, 2004: "Two years from now, spam will be solved."

* Foreword to the OS/2 Programmer's Guide, 1987: "I believe OS/2 is destined to be the most important operating system, and possibly program, of all time."

* COMDEX keynote speech, 2002: "Within five years, I predict it (Windows Tablet) will be the most popular form of PC sold in America."

* "By the time you read this story, the quirky cult company (Apple)…will end its wild ride as an independent enterprise."—Fortune, February  19, 1996

* "Apple [is] a chaotic mess without a strategic vision and certainly no future."—TIME, February 5, 1996

* "Whether they stand alone or are acquired, Apple as we know it is cooked. It's so classic. It's so sad."—A Forrester Research analyst, January 25, 1996 (quoted in The New York Times)

* "The NeXT purchase is too little too late. Apple is already dead."—Nathan Myhrvold (Microsoft's chief technology officer, June 1997)

* "Apple's erratic performance has given it the reputation on Wall Street of a stock a long-term investor would probably avoid."—Fortune, February 19, 1996

* "For all of his success, all Steve Jobs had really accomplished was a temporary pause in Apple's long-term decline."—Infinite Loop, 1996, by Michael S. Malone

* "I'd shut [Apple] down and give the money back to the shareholders."—Michael Dell, founder and CEO of Dell, Inc., 1997

So far, I’ve not seen one single prediction about entirely new products or industries  AI will create. Everything is “AI-assisted X,” as once upon a time everything was “Internet X” or X.com. 

That will not be helpful, even if a fairly reasonable description of how early AI (AI-assisted existing processes and products) will be applied. 

What many of us are looking for are the unexpected and brand-new use cases, products, applications and revenue sources AI will create. 

Right now, our crystal balls are not useful.