Who Wins Next from Generative AI?

In the early stages of a technology change cycle, the firms that benefit the most are those that provide the enabling platform. These are the firms that develop the basic technologies that other firms will use to build new products and services. 

In the generative AI business, that means Nvidia selling graphics process units. That fits an older pattern of new technology opportunities, where infra has to be built first, before use cases, apps and industries can develop. 

For example, in the early days of the internet, the firms that made the most money were the ones that built the infrastructure, such as the internet service providers and the web hosting companies. These firms provided the platform upon which other firms could build their businesses.

As the technology matures, the firms that benefit the most are those that develop applications of the technology. These are the firms that use the technology to create new products and services that meet the needs of consumers. 

For example, in the early days of the internet, once the “plumbing” was in place, the firms that made the most money were the ones that developed e-commerce websites and online advertising platforms.

Only after some time was it possible for substantial new industries and revenue streams to develop. A good road system was required before a mass market auto industry could develop, for example. 

Enabling Infrastructure	Subsequent App and Use Case Development
Roads	Suburbs, car culture, trucking industry
Airports	Passenger airlines, cargo flights, air travel
Power grids	Computers, appliances, electronic devices
Internet	Email, social media, online shopping, cloud computing
Cloud computing	Large-scale applications, such as Netflix and Amazon
Broadband internet	Netflix
Smartphones	Social media
High-speed computing	Generative AI, machine learning, artificial intelligence
Large amounts of data	Natural language processing, computer vision, medical diagnosis
Open standards	Interoperability between systems, collaboration between developers

In perhaps the same way, the early money to be made in generative AI will be reaped by Nvidia and GPU suppliers. Cloud computing will follow fairly closely, as more of the demand for cloud computing services shifts to support of training models and generating inferences. 

Over a period of time, generative AI will reshape consumer and business software, be incorporated into devices and then reshape business processes on a wider scale. 

The bottom line is that suppliers of GPUs have been the first to see significant revenue impact from generative AI. It is easy to predict that cloud computing “as a service” suppliers will see impact next, as well as data centers hosting such operations. 

It will take much longer for existing firms and processes to incorporate generative AI in ways that produce significant financial outcomes, as most of those applications will be indirect: gen AI will be used by existing processes. 

In most cases, applied gen AI will augment or improve existing processes, but without clearly-measurable financial impact. In the medium term, we should start to see some measurable changes as functions are revamped, replacing some legacy methods with gen AI replacements. 

Only eventually will whole new industries arise, supplying novel products. 

Will IoT Revenues Move the Needle for Telcos or Mobile Operators?

It likely still is too early to know whether explosive revenue growth for mobile and telco connectivity providers will be produced by internet of things use cases. 

A large part of the issue is that even exponential growth of IoT devices is not the same thing as direct connectivity using mobile or fixed networks. Perhaps most of the connections will use wireless networks such as Wi-Fi, Bluetooth or other alternatives. 

source: IoT Analytics 

So “devices” in use does not necessarily tell us very much about how markets for IoT connectivity provided by mobile operators or telcos could develop. 

source: IoT Analytics 

All that noted, it is possible that IoT revenues earned by mobile and fixed network operators could be significant by 2030, “significant” using a benchmark of at least $1 billion in annual revenue for any single supplier. 

“Connectivity” will continue to be the big revenue contributor, though some amount of system integration or bundled device sales also will be part of the overall revenue stream. 

Year	Connectivity Revenue ($billion)	System Integration Revenue ($billion)	Device Sales Revenue ($billion)
2022	10.8	2.5	1.5
2023	12.0	3.0	1.7
2024	13.5	3.5	2.0
2025	15.0	4.0	2.2
2026	16.5	4.5	2.5
2027	18.0	5.0	2.7
2028	19.5	5.5	3.0
2029	21.0	6.0	3.2
2030	22.5	6.5	3.5

By way of comparison, if global service provider revenue is at least $1.5 trillion, then IoT revenues might represent about one percent of total revenues in 2022. If global revenue grows about two percent per year to 2030, then 2030 global revenue will be $1.76 trillion. 

Then IoT might represent nearly two percent of total revenue. That of course is based on a far-higher expected growth rate for IoT revenues, compared to total industry revenue growth. 

Global Region	2021 Revenue	2022 Revenue	22/21 Growth
Americas	$572	$580	1.4%
Asia/Pacific	$467	$481	3.0%
EMEA	$438	$449	2.4%
Grand Total	$1,478	$1,510	2.2%

Of course, skeptics or realists might note that, by 2030, mobile operator and telco revenue sources will still rely mostly on mobile subscriptions (60 percent of total revenue) and home broadband (20 percent of total revenue), with all voice services driving about 15 percent of revenue and everything else at about five percent.

Revenue Source	Percent of Total Revenue
Consumer mobile revenues	35%
Business mobile revenues	25%
Home broadband	20%
Voice services	15%
Other revenue sources (IoT, cloud, and enterprise services)	5%

The point, of course, is that even if IoT revenues grow as many expect, they are not likely to drive mobile or telco financial performance. Market share or average revenue per account changes are going to have the greatest impact for individual firms, in nearly all cases.

Watch Out for Irrational Exuberance

Veterans of the telecom industry have lived through financial bubbles before. That is what happens when greed overcomes rational fears about over-investment.

As much as internet service, data transport providers and, in some cases, data center operators, must grapple with ever-increasing demand, those industries also have seen bouts of over-investment in capacity. Think about long-haul fiber transport in the late 1990s or excessive prices paid to acquire 3G spectrum in Europe. 

Excessive optimism led to mispricing of assets, over-investment, questionable growth metrics and outright fraud. Not to mention use of leverage. 

Asset-backed bonds played a role in the telecom collapse of 2001. During the late 1990s, there was a frenzy of investment in the telecommunications industry. Many telecom companies borrowed heavily to finance their expansion, and they often used asset-backed bonds to do so. These bonds were backed by the future revenue streams of the telecom companies, but when the telecom bubble burst in 2000, the value of these bonds plummeted. This led to a number of telecom companies defaulting on their debt, which in turn caused the collapse of the industry.

Merchant banking and excessive use of “junk bonds” also played a role. Telecom companies used merchant banking to finance their expansion by issuing high-yield debt securities, also known as junk bonds. These bonds were often sold to unsophisticated investors who were attracted by the high yields.

Synthetic securitization also increased risk., Synthetic securitization is a type of securitization that involves creating a security that is linked to the performance of another security. Telecom companies used synthetic securitization to transfer their debt to other investors. This allowed them to reduce their debt burden, but it also made them more vulnerable to default if the underlying assets declined in value, which they did. 

Off-balance sheet financing also minimized the actual amount of debt firms were taking on. 

Even debt-equity swaps increased risk. Telecom companies used debt-equity swaps to reduce their debt burden, but it also made them more vulnerable to default if their stock price declined.

Aggressive accounting practices also used aggressive accounting practices to inflate their profits. In some cases, such practices drifted into fraud. Think Worldcom and Enron. 

These risky financing schemes allowed telecom companies to borrow more money than they could afford.

It is the sort of “irrational exuberance” one often has to watch out for.

Streaming Actually Has Higher Costs than Did Linear, Fewer Revenue Sources

The video streaming business, going direct to consumer, has higher costs than the older linear video subscription model for the same reason other retail distribution models come with higher costs than wholesale models. 

Wholesale is a business-to-business transaction with a few customers. Retail is a business-to-consumer transaction with many to millions of customers. Wholesale avoids the costs associated with selling to actual end users, including marketing, stocking, point-of-sale operations and handling of returns, plus lots more customer service and billing and payments processes. 

Cost element	Video streaming	Linear video
Content	Higher	Lower
Marketing	Higher	Lower
Technology	Higher	Lower
Operations	Lower	Higher
Total costs	Higher	Lower
Revenue	Subscriptions	Advertising

Beyond that, revenue sources generally change. Where linear TV has dual revenue streams (advertising and subscription fees), video streaming has generally featured a reliance on subscriptions mostly, though in recent days we have seen a proliferation of ad-supported services as well. 

In addition, linear TV subscription revenue models have included affiliate fee payments to content owners from distributors. Streaming generally has not featured that revenue stream. 

Also, streaming--like any other internet-based business--hinges on scale. For streaming services, that means “going global” rather than remaining a “national” service. 

Still, operating as a “direct to consumer” provider means supporting retail costs in the marketing area that did not exist in the linear format. 

Cost	Linear	DTC
Acquisition marketing	$5-$10 per subscriber	$10-$15 per subscriber
Retention marketing	$1-$2 per subscriber	$2-$3 per subscriber
Avoided expenses	$2-$3 per subscriber	$0
Total marketing costs	$8-$13 per subscriber	$12-$18 per subscriber