Tuesday, June 4, 2024

Where is the AI Edge?

Artificial intelligence processing “at the edge” most likely needs to be qualified, as most of the processing is likely to happen on devices such as smartphones and PCs. Mizuho analysts, for example, forecast one billion AI smartphones shipped from 2024 to 2027. Intel, for its part, expects to ship 40 million AI PC processors in 2024 alone. 

 

Year

On-Device AI Chip Shipments (Millions)

On-Device AI Deployments (Billions)

On-Device AI Sales ($US Billion)

Source

2024

5.2 - 6.0

12.0 - 14.4

28.5 - 34.2

IDC AIoT Market Forecast 

2025

6.8 - 8.0

16.3 - 19.2

42.1 - 49.8

Gartner Forecast: Edge Computing

2026

8.7 - 10.2

21.0 - 25.2

62.4 - 74.7

Digi-Capital [On-Device AI Report]

2027

11.0 - 13.0

27.0 - 32.4

89.1 - 105.3

Yole Developpement [AI Hardware Market Report


Keep in mind that “AI capabilities” are going to be a feature of PCs and smartphones, rather than a distinct product category. Irrespective of the ultimate value of AI on such devices, we will be able to measure sales of products that are AI-capable, even if we might not easily be able to measure the incremental usefulness of AI on PCs and smartphones. 


Year

AI PC Sales Revenue

AI Smartphone Sales Revenue

Source

2024

102.3 - 127.8

214.6 - 268.3

Gartner

2025

148.5 - 185.7

287.1 - 352.4

IDC

2026

207.2 - 259.8

382.4 - 470.9

Counterpoint Research

2027

281.4 - 347.1

501.2 - 613.8

Strategy Analytics


AI “as a service” revenues might be robust as well. 


Year

Gartner

Forrester

IDC

McKinsey

Average

2024

132.5

117.2

125.8

150.4

128.9

2025

172.8

154.1

168.3

192.7

171.9

2026

223.7

202.4

221.5

247.8

221.1

2027

288.2

262.3

285.4

315.7

287.9

Use of Unstructured Data Might Give AI-Driven Financial Advice an Edge

The use of artificial intelligence to provide financial advice has far to go, at least partly for reasons of human trust in the quality of the advice. But there are some potential advantages for AI-generated algorithms that underpin the advice, when compared to current human-devised algorithms. 


It is relatively hard for human algorithms to quantify and make sense of unstructured data. That should be increasingly possible for AI, which will have the advantage of many more data points. On the other hand, AI systems might not be as well-equipped as human advisors when it comes to factoring in human emotions (fear versus greed). 


The same arguably holds true for unexpected events such as pandemics, wars or political upheavals. As a general rule, then, AI-generated advice might be more useful during times of relative market stability, while human decision-making might be more important at times of market turbulence. 


On the other hand, since AI can process so many more sources of data, faster than the human algorithms, it is possible AI also could be better at detecting shifts of trend. 


Also, to the extent that the AI algorithms can evolve organically over time, as the systems learn, AI could have an advantage over human-created algorithms that must be manually revised. For instance, “generally accepted accounting practices” and financial metrics of performance are relatively static measures. 


AI can take advantage of social media sentiment, easy measurement of cars parked in company parking lots and other non-traditional indicators that are similar to “channel checks.”


The point is that we might ultimately be surprised at how much "advice" businesses and functions will eventually be displaced by AI mechanisms. Much expert advice is driven by rules of thumb; general rules and experience or precedent.


AI is going to be pretty good at summarizing and applying such knowledge or precedent for financial guidance. Algorithms already have been driving buy-sell actions in equtiy markets, for example.


In the same way, the ability to process unstructured data might give AI systems an edge over human advice, or at least become a required part of humans giving advice.


Home Broadband Prices have Fallen, But People Don't Believe It

Based on the Producer Price Index (PPI), both home broadband and mobile broadband prices

have fallen consistently since 2016, while prices for electricity, natural gas and postal services have climbed.


The opposite claim--that home broadband prices have risen sharply--often is asserted. So why does it seem as though prices have increased? Because people are buying different products than they once did.

source: Phoenix Center


People once purchased dial-up connections operating at 56 kbps. Now they can buy connections operating at 2 Gbps, 5 Gbps or more. Or they can buy services operating at 100 Mbps to 1 Gbps. 

The faster services cost more, so people spend more. But on an like-compared-to-like basis, recurring prices actually have declined. I used to buy a 56  kbps service for about $20 a month; a 756 kbps service for more than $300 a month. More recently I have bought 300 Mbps or 500 Mbps or 600 Mbps services for prices ranging from $50 to $90 a month (plus assorted taxes and fees). 

My 1-Gbps asymmetrical service (plus taxes, fees, equipment rental, unlimited usage) topped $120. My latest symmetrical gig connection costs less than $90, everything included (taxes, fees, unlimited usage). 

One can argue I am "paying more" than I used to, but what I buy has changed. My speeds are four magnitudes of order higher than they used to be, and prices are less than an order of magnitude higher. Ignoring inlfation effects, one can argue I am "paying more" for home broadband. 

But home broadband comes in different tiers, and I do not buy the same tiers I once did, as I buy different consumer products than I once did. In some cases I pay less; in other cases I pay more, based on the "quality" or "features" of those products. 

And that might help explain why the "home broadband prices are higher" argument seems to make sense. 

One nuance also could be that the PPI measures selling prices for domestic goods and services, essentially tracking inflation at the wholesale level.

So some might argue that the Consumer Price Index (CPI) is the better gauge of what consumers are spending, but since 2015 home broadband prices for speed tiers other than the slowest 10 Mbps downstream product have declined, with the biggest declines coming in the faster speed tiers that most consumers are buying. 


 source: Tech Policy Institute 


All that noted, the reason the “home broadband prices have risen a lot” argument seems to resonate at times is that people are buying different products than they used to. Economists call changes in product quality “hedonic,” meaning that existing products can offer more value over time. 


Hedonic change has been a major trend for electronics products for many decades. 


Product Category

1980s Characteristics

2020s Characteristics

Hedonic Change

Televisions

Bulky CRT displays - Limited screen sizes (under 30 inches) - Standard definition (SD) resolution

Flat-panel displays (LCD, OLED) - Large screen sizes (over 65 inches) - High definition (HD), Ultra High Definition (UHD) or 4K resolution - Smart features with internet connectivity

Increased picture quality, size, functionality

Automobiles

Lower fuel efficiency - Fewer safety features (airbags, ABS) - Manual transmissions common

Improved fuel efficiency - Advanced safety features (automatic emergency braking, lane departure warning) - Automatic transmissions dominant - In-car entertainment systems with navigation and internet connectivity

Increased safety, comfort, and technology features

Smartphones

Brick phones with limited functionality (calls, basic games) - No cameras - No internet access

Large touchscreens - High-resolution cameras - Powerful processors for gaming and apps - Constant internet connectivity

Increased functionality, processing power, and connectivity

Home Appliances

Limited automation (manual controls) - Lower energy efficiency - Fewer features

Smart appliances with programmable features and remote control - Energy-efficient designs - Additional features like ice makers and water dispensers in refrigerators

Increased automation, efficiency, and functionality

Home Computers

Large desktop computers - Limited processing power and storage - No internet access for most

Sleek laptops and desktops - Powerful processors and ample storage - Ubiquitous internet access

Increased portability, processing power, and connectivity


Hedonic change refers to a shift in consumer preferences towards products with higher perceived quality, even if the core function of the product remains the same. The implication is that consumers are willing to pay more for these enhancements that bring greater satisfaction or enjoyment. 


And that is why home broadband prices seem to be “rising” when they actually might be “falling.” Customers are substituting higher-value (and higher price) services for those they once purchased.


Friday, May 31, 2024

What is StreamSaver's Business Purpose?

StreamSaver is a new streaming bundle including Peacock, Netflix, and Apple TV that available exclusively to Xfinity customers, in the same way that Xfinity mobile services are only available to Xfinity customers. 


But it might be reasonable to suggest the new bundle will boost Peacock and Apple TV market share by only single digits.


Netflix, Amazon Prime are the clear leaders leaders, with Disney (Disney, Hulu, coming ESPN service) poised to emerge at the top as well. All the others remain well into secondary or tertiary roles, and the new bundle likely cannot move the needle very much.


Of course, Comcast might see the value not so much in Peacock market share growth (Comcast owns Peacock) but in the value of customer acquisition and retention for its overall subscription businesses (mobile service, home broadband, linear TV).


As with Comcast's mobile phone service, the new bundle can be purchased only by existing Comcast customers (home broadband or linear video service). So the effort is primarily to increase the value of Comcast services.


As often is the case with bundles, the value is partly the price. The advantage of product bundles for the end user (aside from the bundles of features) is almost always the lower price. Streamsaver features the ad-supported versions of the services at $15 per month, where the a la carte prices of those services cost between $6 and $10 a month each.


A reasonable question is how many accounts, and how much revenue, that new service could attract.


Perhaps the easiest assumption is that Xfinity customers who already buy Peacock, Netflix and AppleTV+ will be most inclined to switch. Perhaps next most likely to adopt are Xfinity customers who already buy one or two of those services (Netflix as the presumed anchor and leader) and who can be convinced to the other services. 


The other issue is that the new bundle can only be purchased by Xfinity customers buying either home broadband or Xfinity linear video services. And that means the bundle can likely only be considered by people living in roughly half of U.S. homes. 


The potential market issues can be illustrated by Comcast’s similar approach to selling mobile phone service. 


The context is that Comcast will only sell the bundle to its existing fixed network customers. Since the Comcast fixed network might only reach 45 percent of U.S. dwellings, that puts an upward limit on accounts. 


Then the issue is how much market share Comcast can take from the mobile market leaders (Verizon, AT&T, T-Mobile) that collectively have about 97 percent of the installed base of customers. 


In the third quarter of 2023, Comcast’s Xfinity mobile service had about six million accounts, with revenue growth near 25 percent per year. 


According to the CTIA, there are 523 million mobile devices in the U.S. market with 97 percent of adults owning a mobile phone. Using a (definition of “adults” as those 18 or older, the U.S. Census Bureau estimates that in 2020, adults represented 78 percent of the total population, or 258.3 million people.


If 97 percent of those people have mobile phones and service, that implies something on the order of 250.55 million accounts. If correct, Comcast has an installed base of about two percent of the market. 


The U.S. Census Bureau estimates there were around 140.7 million housing units in the United States in 2020, about 90 percent occupied. So assume dwellings with people living in them at about 126.6 million, with an average occupancy of 2.5 people, for a total of 316.58 million potential mobile accounts. Assume 97 percent have such accounts, for an implied subscriber base of around 307 million, excluding separate business accounts. 


So the installed base of accounts might range from 250 million to 307 million. Assume Comcast’s network passed about 57 million homes (including small business accounts, that figure could reach 80 million locations, by some estimates. Assume it continues to sell exclusively to its own customers. Assume that its installed base of customers is about 55 percent of homes passed (customers buying internet access or video services). 


That suggests terminal adoption ranging from 31.4 million to 44 million locations. If each location features 2.5 accounts, that implies a theoretical terminal customer base of perhaps 78.5 million to 110 million accounts. 


Those figures are the theoretical maximum, keeping in mind that the leading mobile service providers today have installed bases in the 30 percent range each. According to Statista, in May 2023, mobile market shares were:

Verizon: 34.9%

AT&T: 32.2%

T-Mobile: 29.5%

Other Carriers: 3.4%. 


Many observers believe Xfinity therefore will not reach terminal share as high as 55 percent, even within its own service territory, as its fixed network reaches perhaps 45 percent of occupied homes. So with the present policies, Comcast cannot sell to 55 percent of the market. 


With those conditions, were Comcast to reach relative parity with any of the leading mobile service providers, it might hope to reach a ceiling of about 14 percent total market share (assuming 30 percent as a reasonable share within its sales territory, representing 45 percent of locations). 


The new streaming bundle will face the same geographic limitations as does mobile service. On the other hand the streaming market still is growing, and is not mature, as is mobile service. And, obviously, the attraction of the bundle--for Peacock and AppleTV--is the potential to increase share in a market dominated by Netflix, Amazon Prime and Disney. 


Streaming Provider

Estimated Market Share

Netflix

32%

Amazon Prime Video

22%

Hulu

14%

Disney+

12%

Other (HBO Max, Apple TV+, Peacock, etc.)

20%


But the bundle also increases the range of products and value for Xfinity customers. Right now, it is hard to say whether ultimate value comes from growing Peacock share or increasing the value of the Xfinity service overall (reduced churn, for example) or serving as a means of supporting higher revenue per account. 


It might be reasonable to assume all three sources of value will be in play. There are lots of moving parts, though, as most of the other leading or contending streaming firms also are moving to create bundles. 


The coming Disney and Warner Bros. Discovery bundle combines Disney+ with Hulu (including both ad-supported and ad-free options) and Warner Bros. Discovery Max content. 


Also, a new sports streaming service featuring content from Disney (through ESPN), Fox, and Warner Bros. Discovery, Venu Sports, also is launching. Unknown at this point is whether the coming ESPN streaming service will be available as part of that bundle. 


The differences are that StreamSaver can only be purchased by Xfinity customers. The other bundles can be bought by any U.S. consumer.  


Still, the streaming bundles seem to suggest that Netflix and Disney could emerge as the tentpoles of the business. 


Most Consumers Want "Good Enough" Home Broadband

Though regulators and advocates often focus mostly on availability (coverage) and quality (speed), consumers often prioritize value , prefer...