Data and Definitions

Posted on

Last week the German Bundeskartellamt (“Federal Cartel Office”) announced in a press release:

The Bundeskartellamt has initiated a proceeding against the technology company Apple to review under competition law its tracking rules and the App Tracking Transparency Framework. In particular, Apple’s rules have raised the initial suspicion of self-preferencing and/or impediment of other companies, which will be examined in the proceeding.

The press release quoted Andreas Mundt, the President of the Bundeskartellamt, who stated:

We welcome business models which use data carefully and give users choice as to how their data are used. A corporation like Apple which is in a position to unilaterally set rules for its ecosystem, in particular for its app store, should make pro-competitive rules. We have reason to doubt that this is the case when we see that Apple’s rules apply to third parties, but not to Apple itself. This would allow Apple to give preference to its own offers or impede other companies.

The press release continues:

Already based on the applicable legislation, and irrespective of Apple’s App Tracking Transparency Framework, all apps have to ask for their users’ consent to track their data. Apple’s rules now also make tracking conditional on the users’ consent to the use and combination of their data in a dialogue popping up when an app not made by Apple is started for the first time, in addition to the already existing dialogue requesting such consent from users. The Identifier for Advertisers, classified as tracking, which is important to the advertising industry and made available by Apple to identify devices, is also subject to this new rule. These rules apparently do not affect Apple when using and combining user data from its own ecosystem. While users can also restrict Apple from using their data for personalised advertising, the Bundeskartellamt’s preliminary findings indicate that Apple is not subject to the new and additional rules of the App Tracking Transparency Framework.

John Gruber disagrees at Daring Fireball:

I think this is a profound misunderstanding of what Apple is doing, and how Apple is benefiting indirectly from ATT. Apple’s privacy and tracking rules do apply to itself. Apple’s own apps don’t show the track-you-across-other-apps permission alert not because Apple has exempted itself but because Apple’s own apps don’t track you across other apps. Apple’s own apps show privacy report cards in the App Store, too…

If you want to argue that Apple engaged in this entire ATT endeavor to benefit its own Search Ads platform, that’s plausible too. But if Apple actually cared more about maximizing Search Ads revenue than it does user privacy, wouldn’t they have just engaged in actual user tracking? The Bundeskartellamt perspective here completely disregards the idea that surveillance advertising is inherently unethical and Apple has studiously avoided it for that reason, despite the fact that it has proven to be wildly profitable for large platforms.

This strikes me as a situation where Gruber — my co-host for Dithering — is right on the details, even as the Bundeskartellamt is right on the overall thrust of the argument. The distinction comes down to definitions.

It’s striking in retrospect how little time Apple spent publicly discussing its App Tracking Transparency (ATT) initiative — a mere 20 seconds at WWDC 2020, wedged in between updates about camera-in-use indicators and privacy labels in the App Store:

Next, let’s talk about tracking. Safari’s Intelligent Tracking Prevention has been really successful on the web, and this year, we wanted to help you with tracking in apps. We believe tracking should always be transparent, and under your control, so moving forward, App Store policy will require apps to ask before tracking you across apps and websites owned by other companies.

These 20 seconds led, 19 months later, to Meta announcing a $10 billion revenue shortfall, the largest but by no means only significant retrenchment in the online advertising space. Not everyone was hurt, though: Google and Amazon, in particular, have seen their share of digital advertising increase, and, as Gruber admitted, Apple has benefited as well; the Financial Times reported last fall:

Apple’s advertising business has more than tripled its market share in the six months after it introduced privacy changes to iPhones that obstructed rivals, including Facebook, from targeting ads at consumers. The in-house business, called Search Ads, offers sponsored slots in the App Store that appear above search results. Users who search for “Snapchat”, for example, might see TikTok as the first result on their screen. Branch, which measures the effectiveness of mobile marketing, said Apple’s in-house business is now responsible for 58 per cent of all iPhone app downloads that result from clicking on an advert. A year ago, its share was 17 per cent.

These numbers, derived as they are from app analytics companies, are certainly fuzzy, but they are the best we have given that Apple doesn’t break out revenue numbers for its advertising business; they are also from last fall, before ATT really began to bite. They also exclude the revenue Apple earns from Google for being the default search engine for Safari, and while Google’s earnings indicate YouTube has suffered from ATT, search has more than made up for it.

I explained in depth why these big companies have benefitted from ATT in February’s Digital Advertising in 2022; I wrote in the context of Amazon specifically:

Amazon also has data on its users, and it is free to collect as much of it as it likes, and leverage it however it wishes when it comes to selling ads. This is because all of Amazon’s data collection, ad targeting, and conversion happen on the same platform — Amazon.com, or the Amazon app. ATT only restricts third party data sharing, which means it doesn’t affect Amazon at all…

That is not to say that ATT didn’t have an effect on Amazon: I noted above that Snap’s business did better than expected in part because its business wasn’t dominated by direct response advertising to the extent that Facebook’s was, and that more advertising money flowed into other types of advertising. This almost certainly made a difference for Amazon as well: one of the most affected areas of Facebook advertising was e-commerce; if you are an e-commerce seller whose Shopify store powered-by Facebook ads was suddenly under-performing thanks to ATT, then the natural response is to shift products and advertising spend to Amazon.

This is where definitions matter. The opening paragraph of Apple’s Advertising & Policy page, housed under the “apple.com/legal” directory, states:

Ads that are delivered by Apple’s advertising platform may appear on the App Store, Apple News, and Stocks. Apple’s advertising platform does not track you, meaning that it does not link user or device data collected from our apps with user or device data collected from third parties for targeted advertising or advertising measurement purposes, and does not share user or device data with data brokers.

I note the URL path for a reason: the second sentence of this paragraph has multiple carefully selected words — and those word choices not only impact the first sentence, but may, soon enough, lead to its expansion. Specifically:

“Meaning”

Apple’s advertising platform does not track you, meaning that it does not link user or device data collected from our apps with user or device data collected from third parties for targeted advertising or advertising measurement purposes, and does not share user or device data with data brokers.

“Tracking” is not a neutral term! My strong suspicion — confirmed by anecdata — is that a lot of the most ardent defenders of Apple’s ATT policy are against targeted advertising as a category, which is to say they are against companies collecting data and using that data to target ads. For these folks I would imagine tracking means exactly that: the collection and use of data to target ads. That certainly seems to align with the definition of “track” from macOS’s built-in dictionary: “Follow the course or trail of (someone or something), typically in order to find them or note their location at various points”.

However, this is not Apple’s definition: tracking is only when data Apple collects is linked with data from third parties for targeted advertising or measurement, or when data is shared/sold to data brokers. In other words, data that Apple collects and uses for advertising is, according to Apple, not tracking; the privacy policy helpfully lays out exactly what that data is (thanks lawyers!):

We create segments, which are groups of people who share similar characteristics, and use these groups for delivering targeted ads. Information about you may be used to determine which segments you’re assigned to, and thus, which ads you receive. To protect your privacy, targeted ads are delivered only if more than 5,000 people meet the targeting criteria.

We may use information such as the following to assign you to segments:

  • Account Information: Your name, address, age, gender, and devices registered to your Apple ID account. Information such as your first name in your Apple ID registration page or salutation in your Apple ID account may be used to derive your gender. You can update your account information on the Apple ID website.

  • Downloads, Purchases & Subscriptions: The music, movies, books, TV shows, and apps you download, as well as any in-app purchases and subscriptions. We don’t allow targeting based on downloads of a specific app or purchases within a specific app (including subscriptions) from the App Store, unless the targeting is done by that app’s developer.

  • Apple News and Stocks: The topics and categories of the stories you read and the publications you follow, subscribe to, or turn on notifications from.

  • Advertising: Your interactions with ads delivered by Apple’s advertising platform.

When selecting which ad to display from multiple ads for which you are eligible, we may use some of the above-mentioned information, as well as your App Store searches and browsing activity, to determine which ad is likely to be most relevant to you. App Store browsing activity includes the content and apps you tap and view while browsing the App Store. This information is aggregated across users so that it does not identify you. We may also use local, on-device processing to select which ad to display, using information stored on your device, such as the apps you frequently open.

Just to put a fine point on this: according to Apple’s definition, collecting demographic information, downloads/purchases/subscriptions, and browsing behavior in Apple’s apps, and using that data to deliver targeted ads, is not tracking, because all of the data is Apple’s (and by extension, neither is Google’s collection and use of data from Safari search results, or Amazon’s collection and use of data from its app; however, a developer associating an in-app purchase with a Facebook ad is).

“And”

Apple’s advertising platform does not track you, meaning that it does not link user or device data collected from our apps with user or device data collected from third parties for targeted advertising or advertising measurement purposes, and does not share user or device data with data brokers.

One thing should be made clear: there has been a lot of bad behavior in the digital ad industry. A particularly vivid example was reported by the Wall Street Journal last month:

The precise movements of millions of users of the gay-dating app Grindr were collected from a digital advertising network and made available for sale, according to people familiar with the matter. The information was available for sale since at least 2017, and historical data may still be obtainable, the people said. Grindr two years ago cut off the flow of location data to any ad networks, ending the possibility of such data collection today, the company said.

The commercial availability of the personal information, which hasn’t been previously reported, illustrates the thriving market for at-times intimate details about users that can be harvested from mobile devices. A U.S. Catholic official last year was outed as a Grindr user in a high-profile incident that involved analysis of similar data. National-security officials have also indicated concern about the issue: The Grindr data were used as part of a demonstration for various U.S. government agencies about the intelligence risks from commercially available information, according to a person who was involved in the presentation.

Clients of a mobile-advertising company have for years been able to purchase bulk phone-movement data that included many Grindr users, said people familiar with the matter. The data didn’t contain personal information such as names or phone numbers. But the Grindr data were in some cases detailed enough to infer things like romantic encounters between specific users based on their device’s proximity to one another, as well as identify clues to people’s identities such as their workplaces and home addresses based on their patterns, habits and routines, people familiar with the data said.

It’s difficult to defend any aspect of this, and this isn’t even a worst case scenario: there are plenty of unscrupulous apps and ad networks that include explicit Personal Identifiable Information (PII) in these data sales/transfers as well; as Eric Suefert noted in 2020, the industry has had this reckoning coming for a very long time.

That, though, is why the “and” from Apple is so meaningful; here is the sentence again:

Apple’s advertising platform does not track you, meaning that it does not link user or device data collected from our apps with user or device data collected from third parties for targeted advertising or advertising measurement purposes, and does not share user or device data with data brokers.

This definition conflates two very different things: linking and sharing. The distinction between the two undergirded a regular feature of Meta CEO Mark Zuckerberg’s appearances in Congressional hearings; here is a representative exchange between Senator Edward Markey and Zuckerberg in 2018:

Should Facebook get clear permission from users before selling or sharing sensitive information about your health, your finances, your relationships? Should you have to get their permission?…

Senator…I want to be clear: we don’t sell information. So regardless of whether we get permission to do that, that’s just not a thing we’re going to do.

Meta doesn’t sell data; it collects it, and the third parties that leverage the company’s platforms for advertising very much prefer it that way. PII is like radioactive material: it’s very valuable, and can certainly be leveraged, but it’s also difficult to handle and can be dangerous to not just the users identified but to the companies holding it. The way Meta works is that its collective advertising base has effectively deputized the company to collect data on their behalf; that data is not exposed directly, but is instead used to deliver targeted advertisements that are by-and-large bought not by targeting specific criteria, but rather by specifying desired results: app installs, e-commerce conversions, etc. Everything user-related is, to the companies buying the ads, a complete black box.

This is where linking comes in: apps or websites that leverage Facebook advertising (or any other relevant advertising platform, like Snap) include a Facebook SDK or Pixel that tracks installs, sales, etc., and sends that data to Meta where it can be linked to an ad that was shown to that user. Again, this is completely invisible to the developer or merchant; technically they are sending data to Meta, since the conversion data was collected in their app or on their website, but in reality it is Meta collecting that data and sending it to themselves.

The reason why developers and merchants are happy with this arrangement is that advertising is a scale business: you need a lot of data and a lot of customers to make targeted advertising work, and no single developer or website has as much scale as, say, a Google or an Amazon; Meta et al enable all of these smaller developers and merchants to effectively work together without having to know each other, or share data.

Google, Amazon, and Facebook's ad businesses operate similarly, but only Facebook is affected by ATT

You can, to be clear, object to this arrangement, but it’s worth pointing out that this is very different than selling or sharing data with data brokers; all of the data is in one place and one place only, which is broadly similar to the situation with Google or Amazon (or Apple, as I’ll get to in a moment). The big difference is that Meta doesn’t own all of the customer touch points: whereas a Meta advertiser may own their own Shopify website, an Amazon advertiser has to list their goods on Amazon’s site, with all of the loss of control that entails. Apple’s definition, though, lumps Meta’s approach (which again, is representative of other platforms like Snap) in with the worst actors in the space.

“Our”

Apple’s advertising platform does not track you, meaning that it does not link user or device data collected from our apps with user or device data collected from third parties for targeted advertising or advertising measurement purposes, and does not share user or device data with data brokers.

To the extent you think that the Bundeskartellamt is right, then it is this word that is the most problematic definition of all. One would assume that “our” means Apple-created apps, like News or Stocks: just as Amazon collects data from the Amazon app, of course Apple collects data from its own apps. The actual definition, though, is much more expansive; go back to the Epic trial and the exchange I recounted in App Store Arguments:

The argument that Judge Gonzales Rogers seemed the most interested in pursuing was one that Epic de-emphasized: Apple’s anti-steering provisions which prevent an app from telling a customer that they can go elsewhere to make a purchase. Apple’s argument, in this case presented by Cook, goes like this:

A tweet from Adi Robertson

This analogy doesn’t work for all kinds of reasons; Apple’s ban is like Best Buy not allowing products in the store to have a website listed in the instruction manual that happens to sell the same products. In fact, as Nilay Patel noted, Apple does exactly this!

A tweet from Nilay Patel

The point of this Article, though, is not necessarily to refute arguments, but rather to highlight them, and for me this was the most illuminating part of this case. The only way that this analogy makes sense is if Apple believes that it owns every app on the iPhone, or, to be more precise, that the iPhone is the store, and apps in the store can never leave.

Let me be precise in a different way that is relevant to this Article; Apple doesn’t particularly care about or claim ownership of the content of an app on the iPhone, but:

  • Apple insists that every app on the iPhone use its payment system for digital content
  • Apple treats all transactions made through its payment system as Apple data
  • Ergo, all transactions for digital content on the iPhone are Apple data

The end result looks something like this — i.e. strikingly similar to Facebook, but with App Store payments attached:

Apple's ad model looks similar to Facebook's

Here’s the key point: when it comes to digital advertising, particularly for the games that make up the vast majority of the app advertising industry, transaction data is all that matters. All of the data that any platform collects, whether that be Meta, Snap, Google, etc. is insignificant compared to whether or not a specific ad led to a specific purchase, not just in direct response to said ad, but also over the lifetime of the consumer’s usage of said app. That is the data that Apple cut off with ATT (by barring developers from linking it to their ad spend), and it is the same data that Apple has declared is their own first party data, and thus not subject to its ban on “tracking.”

This, needless to say, is where legitimate questions about self-preferencing come to the forefront. Developers who want to link conversion data with Facebook are banned from doing so, while they have no choice but to share that data with Apple because Apple controls app installation via the App Store; this strikes me as a clear example of the President of the Bundeskartellamt’s claim that “Apple’s rules apply to third parties, but not to Apple itself”.

I have been very clear that I disagree with those who want to ban all targeted advertising; I believe that targeted advertising is an essential ingredient in a new Internet economy that provides opportunities to small businesses serving niches that are only viable when the world is your market. After all, people who might love your product need some way to know that your product exists, and what is compelling about platforms like Facebook is that it completely leveled the advertising playing field: suddenly small businesses had the same tools and opportunities to advertise as the biggest companies in the world. At the same time, I understand and acknowledge those who disagree with the concept on principle.

What is frustrating about the debate about ATT, though, is that Apple presents itself as a representative of the latter, with its constant declarations that privacy is a human right, and advertisements that lean heavily into the (truly problematic) world of data brokering, even as it builds its own targeting advertising business. Gruber asked me on this morning’s episode of Dithering whether or not I would feel better about ATT if Apple weren’t itself doing targeted advertising, and the answer is yes: I would still be disappointed about the impact on the Internet economy, but at least it wouldn’t be so blatantly anti-competitive.

Apple, to its credit, has made moves that very much align with its privacy rhetoric by cleaning up some of the worst abuses by apps, including significantly fine-tuning location permissions, providing a new weather framework that makes it significantly cheaper to build a weather app (reducing the incentive to monetize by selling location data), and increasing transparency around data collection. Moreover, at this year’s WWDC the company introduced significant enhancements to SKAdNetwork that should make it easier for developers and platforms like Facebook to re-build their advertising capabilities.

At the same time, an increasing number of signals suggest that Apple is set to significantly expand their own advertising business; an obvious product to build would be an ad network that runs in apps (given that these apps run on the iPhone, Apple would in this scenario claim that collecting data about who saw what ad would be first party data, just like transactions are). Yes, Apple tried and failed to build an ad network previously, but a big reason that effort failed is because Apple didn’t collect the sort of data necessary to make it succeed.

What has changed is not just Apple, but also the data that matters: when iAd launched in 2010, digital advertising ran like people still think it does, leveraging relatively broad demographic categories and contextual information to show a hopefully relevant ad;1 what matters today is linking an ad to a transaction, and Apple has positioned itself to have perfect knowledge of both, even as it denies others the same opportunity.

  1. This is the era when Facebook earned its reputation for being far too cavalier with user data; Facebook was also the company that built the modern advertising approach that depends on linking data instead of sharing it. ↩

Zero-COVID and Free Speech

Posted on

From the Financial Times, last Thursday:

Barely a week after the Chinese Communist party declared victory in its struggle to protect Shanghai from coronavirus, half of the financial hub’s districts will be shuttered this weekend to test millions of residents after signs emerged of renewed community transmission of the virus. China’s most populous city, which was only released from a two-month lockdown last week, detected 11 new infections on Thursday, six outside the city’s mass quarantine centres. The measures will affect eight of the financial hub’s 16 districts, including Pudong, one of the worst-hit areas at the start of the lockdown.

Three cases were detected in the Red Rose beauty parlour in the city centre, prompting health authorities to test more than 90,000 people close to the salon. Only a few days previously, the Xuhui local party body wrote a celebratory post on the microblogging platform Weibo hailing the salon’s reopening on June 1 for clients who had gone weeks without a professional haircut. It said the state-run salon’s resumption of business reflected how the city’s “pandemic situation improved”. The post has since been taken down.

The mass testing ended up finding 5 cases; Chaoyang district in eastern Beijing, meanwhile, is undergoing mass testing of its own, and schools are closed.

One of the common responses to China’s draconian efforts to control COVID’s spread (which, notably, do not include forced vaccination, or the use of Western vaccines), is that it doesn’t work: SARS-CoV-2, particularly the Omicron variant, is simply too viral. It’s worth pointing out that this response is incorrect: China not only eventually controlled the Wuhan outbreak, and not only kept SARS-CoV-2 out for most of 2021, but also ultimately controlled the Shanghai outbreak as well. The fact there were only 5 community cases over the weekend is proof that China’s approach works!

What I think people saying this mean is something different: either they believe the trade-offs entailed in this effort are not worth it, or they simply can’t imagine a government locking people in their homes for months, hauling citizens off to centralized quarantine, separating parents and children, entering and spraying their homes, and killing their pets. I suspect the latter is more common, at least amongst most Westerners: people are so used to a baseline of individual freedom and autonomy that the very possibility of the reality of COVID in China simply does not compute.

Taiwan and Zero-COVID

Perhaps it is not only my knowledge of China, but also my experience living in Taiwan for nearly two decades, or more pertinently, my experience of living in Taiwan the last two years, that makes me much more willing to believe in the effectiveness of China’s approach.

For most of the last two-and-a-half years Taiwan was COVID-free; for most of 2020 that meant life went on as normal, with no masks, everything open, etc.; the one abnormality was that every person entering Taiwan had to quarantine (at home or in a hotel) for 14 days. Things changed in 2021, when the Alpha variant broke through, leading to a soft lockdown: restaurants and schools were closed, and workplaces were strongly encouraged to work from home; masks were instituted everywhere, including outside, and quarantine was hotel only. What is less known is that quarantine went beyond travelers: anyone who was a close contact of an infected person, including family members and co-workers, but also people who might have had the misfortune of being in the same restaurant at the same time as a positive case, were quarantined as well (your location in said restaurant was ascertained by reviewing your cellular location data).

It is this last point that, in my estimation, stopped the 2021 spread in its tracks, and kept Taiwan COVID-free until earlier this year (I myself endured an 18-day centralized quarantine due to testing positive at the airport). It is also, for nearly every Westerner I have relayed this fact to, a startling abridgment of civil liberties. The very idea that you can be locked up for simply being in the wrong place at the wrong time is inconceivable; that, though, is much less stringent than China’s approach in Shanghai, including the requirement that you need a PCR-test within the last 72 hours to even grocery shop.

Here’s the thing: that relative reduction in stringency relative to China is precisely why Taiwan’s containment eventually failed; Taiwan, for most of the last month, has had the highest case rate in the world. From the New York Times COVID tracker:1

NYT Covid tracker on June 13 shows that Taiwan has the highest case rate in the world

Taiwan, to its credit, did not lockdown in the face of this outbreak; I suspect the horrors of the Shanghai lockdown served as a deterrent, particularly given Taiwan’s ongoing struggle for international recognition and desire to distinguish itself from China. It’s also worth noting that at the critical moment — late March and early April — it wasn’t clear if China’s lockdowns would work; still, even if the outcome was clear, Taiwan — despite its willingness to violate civil liberties to a considerably greater degree than most Western democracies — was never willing to go as far as China. And so, while the Chinese approach worked, it almost certainly would not have worked in Taiwan simply because the latter wasn’t willing to be as brutal as the former.

I am being, as best as I can, impartial about the choices here: the important takeaway is not simply that China’s approach did in fact work to arrest the spread of SARS-CoV-2, but also that it was the only approach that worked; even Taiwan’s approach, which was far more stringent than any Western country would tolerate, eventually failed. Of course there were benefits, particularly in terms of getting time to administer vaccines, but it’s certainly worth wondering if it was all worth it.2

The opposite side of the spectrum were areas of America that, after enduring a few months of (very) soft lockdown at the beginning of the pandemic, were mostly open from the summer of 2020 on; I have friends in parts of Wisconsin, for example, whose kids have been in school since the fall of 2020. The price of this approach was far more deaths, particularly amongst the elderly who have always been at far higher risk: over 1 million Americans have died of COVID.

This isn’t the complete COVID story, though, and not simply because there can be no honest accounting of the pandemic until it finally sweeps China; the most effective vaccines in the world were developed in the West, and the U.S. produced and distributed the largest number of them. How many lives were saved, and how much economic upheaval — which isn’t about simply dollars and cents, but people’s livelihoods, sense of worth, and even sanity — was avoided or reduced because of vaccines? That must be recorded in the ledger as well, and in this accounting the West comes out looking far stronger.

The Great Firewall

The reason to audit this accounting is that I think there is an analogy to be drawn between COVID and the debates around free speech that have sprung up over the last six years. Before then, there wasn’t much of a debate about free speech: just as the W.H.O. and C.D.C. used to maintain that lockdowns don’t work, it used to be widely accepted that free speech was a good thing. Moreover, it was also accepted that free speech was not simply a legalistic limitation on government power, but was a cultural value. I pointed out earlier this year that this was no longer the case in elite culture; the debate around Elon Musk buying Twitter confirmed exactly that.

To summarize, the “sophisticated” view on free speech is that the First Amendment both restricts the government and also protects companies who make their own moderation decisions; this is of course correct legally, but the idea that this distinction should be both celebrated and pushed to its limit is new. That, by extension, means that the “rube” view on free speech is that said principle ought to apply broadly: not only should the government not be able to limit your speech, but neither should Facebook or Twitter or Google. Again, this was a widely held view not too long ago: much of the debate around net neutrality, for example, centered on the importance of private corporations not being allowed to treat different bits of data differently based on what type of content they represented.

There are, of course, philosophical arguments to be made as to why either view is better or worse than the other; to return to the COVID analogy, one can debate whether or not the sacrifice of civil liberties is worth whatever deaths might be prevented (again, with the caveat that the final accounting is not yet complete). What I think is missing in both debates, though, is the question of what was possible.

Go back to my point above: I strongly suspect that most people in the West are convinced that China’s approach will not work — even though it is! — because they simply cannot imagine enduring or tolerating or even encountering the level of brutality necessary for success; that is certainly true of COVID dead-enders who still bemoan that the West isn’t doing enough to control the spread of COVID. It is, from my perspective, hard to imagine any of these folks accepting non-negotiable centralized quarantine simply for being in the wrong restaurant at the wrong time — and again, this is the Taiwan approach that ultimately failed! They are complaining about something that simply isn’t possible, not because their political enemies are unwilling to do what is necessary, but because they themselves would never tolerate it.

This, I should note, is why I have long been in strong favor of fully opening up: while there was an argument to be made that it was worth trying to delay outbreaks until vaccines were widely available, by the summer of 2021 (in the U.S.) the only possible outcome of restrictions was to make people miserable at best, and cause economic, socio-political, and developmental damage at worst; spread, absent a China-style approach, was inevitable, so why invite bad outcomes when there are no benefits?3

I have the same questions about free speech. Once again, I must acknowledge that China’s approach to free speech works, at least in terms of its leaders’ immediate goals. In other words, it doesn’t exist, even — especially! — on the Internet. This — like China’s insistence on zero-COVID — was something that Westerners scoffed at as being unrealistic; then-President Bill Clinton said upon the establishment of Permanent Normal Trade Relations with China:

In the new century, liberty will spread by cell phone and cable modem. In the past year, the number of Internet addresses in China has more than quadrupled, from 2 million to 9 million. This year the number is expected to grow to over 20 million. When China joins the W.T.O., by 2005 it will eliminate tariffs on information technology products, making the tools of communication even cheaper, better, and more widely available. We know how much the Internet has changed America, and we are already an open society. Imagine how much it could change China.

Now there’s no question China has been trying to crack down on the Internet. Good luck! That’s sort of like trying to nail jello to the wall. But I would argue to you that their effort to do that just proves how real these changes are and how much they threaten the status quo. It’s not an argument for slowing down the effort to bring China into the world, it’s an argument for accelerating that effort. In the knowledge economy, economic innovation and political empowerment, whether anyone likes it or not, will inevitably go hand in hand.

Clinton, along with nearly all of the Western intelligentsia, underrated China’s willingness to do whatever it took to build a mold around that jello, from building the Great Firewall to employing countless numbers of censors to tanking its entire IT sector once it felt it was becoming too politically powerful. The end result is a populace that not only has little idea about today’s reality — i.e. that most people have had COVID, and are fine, and are living normally — but even less idea about the past.

Tank Cake

Last February Time Magazine named Li Jiaqi one of its “Next Top 100 Most Influential People”. Li’s nickname was the “lipstick king”, which refers to the time in 2018 when the live-streaming e-commerce peddler sold 15,000 lipsticks in 5 minutes; last fall Li sold $1.7 billion worth of goods in 12 hours. Ten days ago, on June 3, Li was doing what he does best — selling goods via live-streaming — when his stream suddenly went off the air; Li, within a matter of hours, was suddenly off of the Internet, no longer appearing on Taobao, Alibaba’s e-commerce platform that streamed his show. The BBC explained what happened:

Last Friday night, Li was mid-way through his popular livestream show when it ended abruptly. The 30-year-old, known for his smooth voice and K-pop idol looks, had just shown his audience a vanilla log cake while selling snacks. The cake resembled a tank: it had Oreos for wheels and a wafer pipe resembling a cannon. And Li’s show was on 3 June, the eve of the 33rd anniversary of the Tiananmen Square massacre…

The "Lipstick King" selling a tank cake on June 3rd

Generations of Chinese have grown up without learning of the massacre – and many of those millennials and Gen Z-ers appeared to be among Li’s audience on Friday and in the days after. Li failed to return to his livestreaming show after the transmission was cut. Shortly after, he posted on his Weibo account saying he had merely faced technical issues. But his continued absence – he has missed three shows so far during one of the year’s biggest online shopping festivals – has only fuelled more questions and debate. Some have cottoned on quickly as to why he was censored, while others are having a revelation. “What does the tank mean?” a confused viewer asked. Another said: “What could possibly be the wrong thing to say while selling snacks?”

That’s not all, though: it seems almost certain that Li had no idea he did anything wrong, or why.

Few online believe that Li was trying to make a political statement. Given his celebrity status, he knew how to navigate political sensitivities and to steer clear of minefields, they said. And he had never expressed political beliefs before. Some even argued that he was possibly among those who didn’t know about the Tiananmen Square massacre.

Many of his loyal fans also wondered if the top livestreamer had been set up by competitors to take a political fall, and perhaps the cake was sneaked into the line-up of his show on Friday. A clip circulating on social media, apparently of the moment before the cake is brought out, also shows Li expressing surprise over the announcement of a tank product. A male assistant announces in the background that the team has a tank-shaped good to sell. Li laughs and says: “What? A tank?” His co-presenter then says: “Let’s see if Li Jiaqi and I will still be here at 11pm.” They were taken off air shortly after 9pm.

Many fans suspect purposeful sabotage; perhaps that is a conspiracy theory, but said theory is undergirded by the reality that it is not just possible but even probable that a 30 year-old in China has no idea that selling a tank-shaped cake on June 3rd is grounds for being disappeared. To put it another way, China’s control of information is not unlike its control of COVID: it seems impossible, and the means intolerable, but that is simply because we in the West can’t imagine the limitations on personal freedom necessary to make it viable.

Acceptance and Competition

To further expand on this point: if people in the West would not accept truly strict lockdowns, then they certainly wouldn’t accept centralized quarantine (which didn’t work), which means they absolutely wouldn’t accept forced testing and the inability to leave your house for months. Ergo, people in the West would never accept the reality of zero-COVID, which is why it makes sense to go in the opposite direction: open up, and forgo the massive costs of zero-COVID as well. Don’t get stuck in the middle, enduring the worst outcomes of both.

Similarly, if people in the U.S. would not accept any government infringement on speech, then they certainly wouldn’t accept ISP-level censorship like the Great Firewall, which means they absolutely wouldn’t accept forced disappearances for selling the wrong cake. Ergo, people in the U.S. would never accept the reality of true control of speech, which is why it makes sense to go in the opposite direction: embrace free speech not just as a law but as a cultural more, and forgo the massive costs of half-ass speech restrictions as well. Don’t get stuck in the middle, enduring the worst outcomes of both.

COVID, alas, seems to have been a worst case scenario in terms of both points: we suffered the aforementioned economic, socio-political, and development damage associated with strict control, while controlling nothing; meanwhile private platforms went overboard in controlling information, and ended up only deepening the suspicion of skeptics about COVID and its vaccines, leading to many more deaths, but also increased skepticism about vaccines generally.

The worry is that this middling approach, where we get the worst of both worlds, impacts innovation generally; China is increasingly focused on a top-down approach to technological innovation in particular, placing heavy emphasis and tons of money on catching up in areas like semiconductors and AI. The best response is to go in the opposite direction, and let a thousand flowers bloom, trusting that innovation by definition arises in places we least expect it.

To put it another way, if we could accurately eliminate bad ideas, then there would, by definition, be no more good ideas to discover; the way to compete with China is to lean into the fact that there remains so much we don’t yet know.

You likely have, by this point, heard the story of Katalin Karikó; from Stat News in 2020:

Before messenger RNA was a multibillion-dollar idea, it was a scientific backwater. And for the Hungarian-born scientist behind a key mRNA discovery, it was a career dead-end. Katalin Karikó spent the 1990s collecting rejections. Her work, attempting to harness the power of mRNA to fight disease, was too far-fetched for government grants, corporate funding, and even support from her own colleagues…By 1995, after six years on the faculty at the University of Pennsylvania, Karikó got demoted. She had been on the path to full professorship, but with no money coming in to support her work on mRNA, her bosses saw no point in pressing on.

Karikó would eventually figure out how to stop the body from rejecting mRNA, an essential discovery on the way to today’s vaccines. Along the way, though, she was nearly defeated by an academic system that increasingly relies on money from the powers that be, who think they know everything; fortunately said powers couldn’t actually stop her work, even though the consensus was that said work was a bad idea.

Only with time did it reveal itself as a good idea, which is the story of almost everything in life: we live, we learn, we discover new things, not just those of us alive in 2022, but all of humanity for our entire existence. That is how we beat COVID: not by destroying our liberties and lives, but by invention and information. It turns out that free speech isn’t just an analogy to COVID: it’s an essential part of getting past it. And, critically, it’s the only approach that nearly all of us reading this article — particularly those of us in the U.S., no matter our political affiliation — would actually tolerate.

In short, we live in the U.S., not China, and it’s high time all of us — including tech companies — started acting like it, instead of LARPing the most pathetic imitation possible.

  1. This case rate is likely significantly underreported, I would add: given that positive cases are not allowed to leave their house for 7 days — again, tracked by cellphone — there is a very strong incentive to simply not report a positive case; anecdotally speaking the majority of people I know in Taiwan have gotten COVID over the last month or so. ↩

  2. My aforementioned 18-day quarantine in April certainly seemed like a needless waste of my life — as do ongoing traveler quarantines whose only purpose is to protect travelers from what is again, the highest case rate in the world. ↩

  3. I do recognize that people wished to wait for a children’s vaccine; given the relative risk for children I disagreed, but I acknowledge the argument ↩

Thin Platforms

Posted on

The Department of Justice’s 1998 complaint against Microsoft accused the company of, amongst other things, tying the Internet Explorer browser to the Windows operating system:

Internet browsers are separate products competing in a separate product market from PC operating systems, and it is efficient to supply the two products separately. Indeed, Microsoft itself has consistently offered, promoted, and distributed its Internet browser as a stand-alone product separate from, and not as a component of, Windows, and intends to continue to do so after the release of Windows 98…

Microsoft’s tying of its Internet browser to its monopoly operating system reduces the ability of customers to choose among competing browser products because it forces OEMs and other purchasers to license or acquire the tied combination whether they want Microsoft’s Internet browser or not. Microsoft’s tying — which it can accomplish because of its monopoly power in Windows — impairs the ability of its browser rivals to compete to have their browsers preinstalled by OEMs on new PCs and thus substantially forecloses those rivals from an important channel of browser distribution.

In retrospect, the complaint feels quaint for three reasons:

First, Microsoft won the browser wars, and it didn’t matter; after peaking at 95% market share in 2004, Internet Explorer was first challenged by Firefox, which peaked at 32% market share in 2010, and then surpassed by Chrome in 2012:

The reasons ended up being both a condemnation and an endorsement of the libertarian defense of Microsoft’s actions, depending on your timeframe: sure, the company leveraged its operating system dominance to gain browser market share, but the company also made a great browser (I personally switched with the release of version 4). And then, with Version 6 and its position seemingly secured, the company just stopped development; that is what opened the door to first Firefox and then Chrome, both of which were downloaded and installed by end users looking for something better. The market worked, eventually.

Of course, the reason the market could work is that Windows was an open platform: sure, Microsoft controlled (and allegedly abused) what could be preinstalled on a new computer, but once said computer was in a user’s hands they could install whatever they wanted to, including alternative browsers. That gets to the second reason why the complaint feels quaint: today having a browser pre-installed is de rigueur for operating systems, and Apple’s iOS goes much further than simply pre-installing Safari: all alternative browsers must use Apple’s built-in rendering engine, which means they can only compete on user interface features, not fundamental functionality.1

The third reason has to do with Microsoft itself.

Thick and Thin

As I noted last week in an Update, one of the overarching themes of CEO Satya Nadella’s Build developer conference keynote was the seemingly eternal tech debate about thin versus thick clients (to dramatically simplify — and run the risk of starting a flame war — thin clients are terminals for a centralized computer, while thick clients are computers in their own right, that sync):

The biggest takeaway from this keynote is that for developers, at least the ones that Microsoft is courting, the thin client model has won — although the truth, as is so often the case with tech holy wars, has ended up somewhere in the middle. Here is the key distinction: there is and will continue to be a lot of work that happens locally; all of the assumptions around that work, though, will be as if the work is being done on the server. For example:

  • GitHub Codespaces is an explicitly online environment that you can temporarily use locally.
  • Azure Arc provides the the Azure control plane for an on-premises development environment.
  • The Azure Container Apps service and Azure Kubernetes Service enable developers to write locally in the same environment they deploy to the cloud.

Moreover, several other of the announcements were about patching up limitations in cloud development relative to local: Microsoft Dev Box, for example, enables the deployment of cloud-based VMs that mimic a local development environment for things like app development; Microsoft Cloud PC (which was previously announced) does the same thing for client applications.

What makes this shift so striking is that it is being articulated by Microsoft; after all, Windows (along with Intel) was the dominant winner of the thick client era. Yes, Windows Server was an integral part of Microsoft’s enterprise dominance, but the foundation of the company’s strategy — as evidenced by the tactics used in the fight against Netscape — was the fact that Windows was the operating system on the devices people used. That, by extension, was precisely why mobile was so disruptive to the company: suddenly Windows was only on some of the devices people used; iOS and Android were on a whole bunch of them as well.

I’ve spent many articles writing about how Satya Nadella weaned Microsoft off of its Windows-centric strategy; the pertinent point in terms of this Article comes from Teams OS and the Slack Social Network:

The end of Windows as the center of Microsoft’s approach, and the shift to the cloud, though, did not mean the end of Microsoft’s focus on integration, or its attempt to be an operating system; the company simply changed its definition of what an operating system was; Satya Nadella said at a press briefing in 2019:

The other effort for us is what we describe as Microsoft 365. What we are trying to do is bring home that notion that it’s about the user, the user is going to have relationships with other users and other people, they’re going to have a bunch of artifacts, their schedules, their projects, their documents, many other things, their to-do’s, and they are going to use a variety of different devices. That’s what Microsoft 365 is all about.

Sometimes I think the new OS is not going to start from the hardware, because the classic OS definition, that Tanenbaum, one of the guys who wrote the book on Operating Systems that I read when I went to school was: “It does two things, it abstracts hardware, and it creates an app model”. Right now the abstraction of hardware has to start by abstracting all of the hardware in your life, so the notion that this is one device is interesting and important, it doesn’t mean the kernel that boots your device just goes away, it still exists, but the point of real relevance I think in our lives is “hey, what’s that abstraction of all the hardware in my life that I use?” – some of it is shared, some of it is personal. And then, what’s the app model for it? How do I write an experience that transcends all of that hardware? And that’s really what our pursuit of Microsoft 365 is all about.

This is where Teams thrives: if you fully commit to the Microsoft ecosystem, one app combines your contacts, conversations, phone calls, access to files, 3rd-party applications, in a way that “just works”…This is what Slack — and Silicon Valley, generally — failed to understand about Microsoft’s competitive advantage: the company doesn’t win just because it bundles, or because it has a superior ground game. By virtue of doing everything, even if mediocrely, the company is providing a whole that is greater than the sum of its parts, particularly for the non-tech workers that are in fact most of the market. Slack may have infused its chat client with love, but chatting is a means to an end, and Microsoft often seems like the only enterprise company that understands that.

Note that line about “3rd-party applications”: if Teams is the Windows of Microsoft’s new services strategy, then it follows that the platform opportunity for developers in Microsoft’s ecosystem is itself centered on Teams; that’s exactly what Nadella described in the Build keynote:

Let’s talk about the future of work and how we’re making apps more contextual and people-centric, so you can build a new class of collaborative applications. It starts with Microsoft Graph, which underlies Microsoft 365 and makes available to you information about people, their relationships, and all of their artifacts. Today we are seeing developers around the world enriching their apps with Microsoft Graph. In fact, more than half of the Microsoft 365 tenant are using custom-built and 3rd-party apps powered by the Graph. With Graph connectors ISVs can extend their applications and have them be discovered as part of the user’s everyday tasks, whether they are writing an email, meeting on Teams, or doing a search. For example, data from an app can appear directly in an organization’s search results, as you can see in the experience Figma is building here. You can compose a mail and @-mention files from these apps in-line, and you can access them in Teams chat too. Another way that you can create interactive experiences is by building live actionable loop component using adaptive cards like partner Zoho does. Your users can make decisions and take action like updating the status of a ticket right in the flow of work, and updates are always live, like this one across Outlook, Teams, and Zoho.

When you combine the Microsoft Graph with Microsoft Teams, you combine the data that describes how people work together with the place they work together. It’s incredibly powerful, and developers are extending their apps into Teams and embedding Teams in their apps. In fact, monthly usage of 3rd-party apps and custom-built solutions on Teams has grown 10x over the last two years, and more and more ISVs are generating millions of [dollars in] revenue from customers using apps built on Teams.

“Graph connectors” are the new APIs.

Windows versus Teams

If the Windows platform looked like this…

The Windows "thick" platform

…then the new Teams platform looks like this:

The Teams "thin" platform

There are a few important observations to make about these differences.

First, in the PC era the monopoly that mattered was being the only operating system on a single device. This, to be clear, is a technical necessity — while a PC could dual-boot into different operating systems, only one could run at a time2 — but it was the foundation of Windows’ market monopoly. After all, whichever operating system was running on the most devices most of the time was the operating system that developers would target; the more developers on a particular operating system, the more popular that operating system would be amongst end users, resulting in a virtuous cycle, aka a two-sided network, aka lock-in, aka a monopoly.

Once mobile came along, though, not only did the number of devices proliferate, but so did the need for new user interfaces, power requirements, hardware re-imagining, etc.; this made it inevitable that Microsoft would miss mobile, because the company was approaching the problem from the completely wrong perspective.3 At the same time, this proliferation of devices meant that the point of integration — which enterprises still craved — moved up the stack. I wrote in 2015’s Redmond and Reality:4

That is because there is in fact a need for an integrated solution on mobile. Look at Box, for example: the company obviously has a cloud component, but they also have multiple apps for every relevant — and non-relevant! — platform resulting in much better functionality than what Microsoft previously had to offer. Multiply that advantage across a whole host of services and it starts to make sense for the CIO to modularize her backend services in order to achieve integration when it comes to how those services are accessed:

A drawing of Pre-Cloud and post-Cloud Services

This is exactly what Microsoft would go on to build with Teams: the beautiful thing about chat is that like any social product it is only as useful as the number of people who are using it, which is to say it only works if it is a monopoly — everyone in the company needs to be on board, and they need to not be using anything else. That, by extension, sets up Teams to play the Windows role, but instead of monopolizing an individual PC, it monopolizes an entire company.

Second, developers had much more power and flexibility in the old model, because they had direct access to the underlying PC. This had both advantages — anyone could make an app that could do anything, and users could install it directly — and disadvantages — anyone could make an app that could do anything, and users could install it directly. In other words, the same openness of the PC that presented an opportunity for Firefox and Chrome to dethrone Internet Explorer — and for Netscape to exist in the first place — also presented an opportunity for viruses, malware, and ransomeware.

This latter point is the justification that Apple returns to repeatedly for its App Store model, even though a significant portion of the increased security of mobile devices is due to fundamentally different architectural choices made in designing the underlying operating system. Then again, these arguments go hand-in-hand: it’s those architectural choices in iOS (and Android) design that make App Store control possible; the broader point is that mobile set the expectation that developer freedom — and by extension, opportunity — would be limited by the operating system owner.

A thin platform like Teams takes this even further, because now developers don’t even have access to the devices, at least in a way that matters to an enterprise (i.e. how useful is an app on your phone that doesn’t connect to the company’s directory, file storage, network, etc.). That means the question isn’t about what system APIs are ruled to be off-limits, but what “connectors” (to use Microsoft’s term) the platform owner deigns to build. In other words, not only did Microsoft build their new operating system as a thin platform, they ended up with far more control than they ever could have achieved with their old thick platform.

Stripe OS

Build wasn’t the only developer conference last week: Stripe also held Stripe Sessions, and one of the tentpole sections of the keynote was called “Finance OS”. Here’s Stripe co-founder and President John Collison:

We’ve talked about payments, and how they’re highly strategic, and rapidly fragmenting, and we’ve talked about the business model innovations of adaptive enterprises and fintech everywhere. These trends are great news for the Internet economy, but a challenge for finance and business operations teams. The rate limiter for so many new opportunities isn’t the idea for a great product; it’s the mundane foundations. “Can we build for this? Can we get international operations off of the ground? Can we expand when we’re still not closing our books on time?” It’s never just about having the idea for a great product, it’s about being able to operate it, and that’s why we’re building a modern operating system for finance, and like any good OS, we’re focused on nailing the basics.

Those basics included features like invoicing, billing, taxes, revenue recognition, and data pipelines, all of which sit on top of the various ways to gather, store, and distribute money that Stripe has abstracted away:

The Stripe OS

This image, given its similarity to the one above, makes clear what was coming next:

So we just heard about core revenue management capabilities, like invoicing, subscription billing, and handling tax. Even if you’re not using Stripe, these are the things you should want running like clockwork.

But what about everything else? It’s like any operating system: core functionality needs to work perfectly out of the box, but the breadth of functionality of the platform is also really important, having an app to solve every use case. For things like customer messaging, you might want to use something like Intercom; for contracts, DocuSign; or, you might just to build your own tool. But often these workflows are highly integrated, so for years our users have been asking us for the tools of their choice to interoperate with Stripe…

We’re thrilled to launch today Stripe Apps and the Stripe App Marketplace, where you can find or build best-of-breed tools that work naturally with Stripe.

There are the missing pieces!

The Stripe thin platform

“Working naturally with Stripe” doesn’t simply mean access to Stripe’s APIs; it means fitting in to the Stripe dashboard — Stripe is even including pre-made UI components so that 3rd-party apps look like they were designed by the fintech company:

Stripe offers pre-made UI components for integrating into the Stripe dashboard

This is another thin platform: developers don’t have access to the core financial data of a company, nor does IT want them to; instead the opportunity is to sit on top of an abstraction layer that covers all of a company’s money-moving pieces, and to fit in as best as you can.

Of course I am covering the Build and Stripe Sessions keynotes together because both happened the same week; at the same time, it was a fortuitous coincidence, because Stripe’s announcement brings important context to Microsoft’s approach. After all, I used the magic word “monopoly”; the truth, though, is that not only was an operating system monopoly inevitable, it also made perfect sense from a user perspective that important functionality — like browsing — became integrated with the core OS.

Collison made the case as to why similar considerations should be front-and-center for thin platforms — there are things “you should want running like clockwork.” Microsoft would make a similar argument about Teams and its incorporation of things like file storage and communications, and, I would argue, Teams’ success in the market relative to Slack is evidence that the argument is a compelling one to customers. That Microsoft has so often seemed like the only enterprise company actually building for an enterprise’s ends, instead of solipsistically obsessing over being best-of-breed for one specific means, seems worth celebrating and emulating, not condemning and complaining.

At the same time, it is also worth mourning the slow eclipse of the thick client model. Yes, things like malware were a pain and a drain on productivity, and the SaaS model has led to a plethora of new products that are accessible to companies without needing an IT department, but the big downside of the thick model in terms of what could go wrong and the necessity of IT created the conditions for massive upside, in this case the opportunity to make new apps — and by extension, new companies — without needing any permission, “connectors” or pre-made UI components. Alas, the tech industry is past the end of the beginning; welcome to middle age, where the only thickness is your waistline.

I wrote a follow-up to this Article in this Daily Update.

  1. Apple argues, not without merit, that this is for security reasons; critics argue, with considerable | merit, that this restricts innovation on iOS and meaningful competition with the App Store. ↩

  2. Absent virtualization, although that wasn’t really feasible on user-level PCs at the time Windows was establishing its dominance ↩

  3. This interview with Tony Fadell includes an excellent discussion on this point in the context of Intel, which applies just as much to Microsoft. ↩

  4. With, I am ashamed to admit, probably the worst drawing in the history of Stratechery; as I recall I was late to a Chinese New Years’ Eve dinner! ↩

Warner Bros. Discovery

Posted on

Last week the Wall Street Journal ran a profile of longtime Discovery CEO (now Warner Bros. Discovery CEO) David Zaslav entitled There’s a New Media Mogul Tearing Up Hollywood, adding “Zas Is Not Particularly Patient”. After opening with an anecdote about Zaslav complaining about Warner Bros. backing a Clint Eastwood movie, even though they thought it would fail, the profile states:

Mr. Zaslav, who last month took over the company resulting from Discovery’s merger with AT&T Inc.’s WarnerMedia, has given every indication he wants to be a talent-friendly mogul, schmoozing with industry personalities at the Beverly Hills Hotel. But the 62-year-old cable-industry veteran, a protégé of the late Jack Welch, longtime CEO of General Electric Co., has shown he isn’t afraid to ruffle the industry’s elite.

He and his team have been scouring the company’s books, making it clear spending needs to be reined in. They have abandoned projects they consider costly and unnecessary. That included pulling the plug on CNN+, barely a month after previous management launched the streaming service, and canceling a DC Comics superhero movie in development. He has given an unwelcome jolt to executives in the WarnerMedia empire who were happy when AT&T decided to part with it in the merger, hoping there would be less financial scrutiny—not more…

Mr. Zaslav has few options other than drastic moves. The deal brought the new company—now home of Warner Bros. and cable channels including HBO, CNN, TNT, Food Network and HGTV—$55 billion in debt, and he has promised to cut at least $3 billion in costs. He has given executives a few weeks to provide restructuring and business plans. “We are not trying to win the direct-to-consumer spending war,” Mr. Zaslav said on an April earnings call. On the call, Chief Financial Officer Gunnar Wiedenfels called out the nearly $30 billion the company spends making and marketing content, saying: “We intend to drive the highest level of financial discipline here.”

The profile continued in the same vein, and came across as fairly negative; that is why I appreciated it, because I am otherwise extremely bullish about the potential for Zaslav’s new company.

HBO and Discovery Synergies

I wrote a year ago when the deal between AT&T and Discovery was announced that Warner Bros. and Discovery had excellent synergies:

Start with the cable bundle: yes, cord-cutting continues, but there are still a lot of households with cable, and this new company will have significantly enhanced bargaining power with distributors. WarnerMedia’s combination of live sports, news, premium television, and scripted shows was already quite strong; Discovery brings a highly differentiated set of channels from HGTV to Discovery to Food Network that not only attract distinct demographics, but also are particularly effective at driving advertising.

Another set of synergies come in the two companies burgeoning direct-to-consumer offerings. Once again the breadth of content is a good thing: HBO Max and Discovery+ have something for everyone in the household. The types of content are complementary as well; back in 2018 I explained in the context of Friends:

While most of the Netflix attention is paid to original series, the truth is that there are two types of shows on the streaming network:

  • Original series drive new subscribers
  • “Filler”, that is, content that is there when subscribers simply want something to watch, keeps people subscribed

Discovery content is excellent filler [while HBO excels at original series].

Zaslav and CFO Gunnar Wiedenfels made all of these points on Warner Bros. Discovery’s inaugural earnings call in late April. Start with the second point above, about streaming synergy. Zaslav, in response to a question about advertising (more on this in a moment), brought up the fact that Discovery+ has very low churn:

We’re in the market already with an ad-light product. We’re the ones that were out there very early saying ad-light looks really compelling, because it’s a great consumer proposition. Our users, the churn was very low; we were doing between two and four minutes of advertising and generating $5, $6 in incremental revenue, and as it scaled, we started to make more. And so, we said very early on, we’re going to switch to offer consumers what they want, a lower-priced opportunity with a small number of advertising.

HBO Max isn’t doing so well in this regard, at least according to Zaslav a few minutes later:

We have some work to do on the platform itself that will be significant. But we also think that one of the big opportunities here is going to be churn reduction. There is meaningful churn on HBO Max, much higher than the churn that we have seen. And so, the ability for us to come together is part of one of the thesis here that managing churn, and we’ve seen this because we’ve been added in Europe for eight years, as we begin to manage churn in a meaningful way, that provides a real meaningful growth.

The benefit of coming together is exactly what I noted a couple of years ago: hits may capture customers, but filler content — especially a wide range of it — keeps them:

What you need is a diversity of content for everybody in the home, and they may come in for Euphoria [from HBO], but our research shows that people watch Euphoria, their favorite second show to watch is 90 Day Fiance [from TLC]. Having a diversity of content is a reason why people are spending hours with Discovery+…when you put all of this diversity of content together, there is content for kids, there is content for teens, it’s basically everybody in the family, why would you go anywhere else. We have all the movies, we have all of the library content that you want…

If you look at HBO right now, what it really needs is precisely what we have. When they are finished with watching Winning Time, they can go and watch Friends or watch Big Bang or watch their favorite movie or go over and watch Oprah or watch some TLC shows just for fun. So, we believe and we see this in Europe where we tried to offer, we thought that the answer was just to offer niche high quality that you get high-quality shock and all content together with a lot of nutrition, in our case in Europe, together with sport and you offer something that everybody in the family uses, and the churn goes way down, it’s much harder to churn out of a product when your kids use it or your significant other uses it or your mom and dad are watching, but also if you find yourself watching it more often. So, I think it’s precisely why we did this deal. And I think everything tells us that it’s going to make us stronger and more compelling because of the breadth of the quality menu of IP that we have.

In short, HBO Max plus Discovery+ is a bundle, with all of the attendant advantages that entails (and which certainly did not apply to a standalone CNN+ service). Of course this also strengthens Warner Bros. Discovery’s hand in terms of the linear TV bundle as well; Zaslav said in his prepared remarks:

One of the company’s unique assets is the linear network group, and in 2021 taken together, we enjoyed the number one share in total television total day in all key demos and people 2+. And we have the greatest brands: HG, Food, HBO, Discovery, CNN, NBA, March Madness, NHL, Magnolia, The Oprah Winfrey Network. Our balanced verticals and content genres across scripted, lifestyle, sports and news provide us with significant opportunities to not only cross-promote for the benefit of the portfolio, but also to offer compelling reach and targeting campaigns for our advertising partners.

Don’t forget negotiating leverage; Wiedenfels noted:

US distribution revenues were up 11% year-over-year, largely driven by the growth of Discovery+ subscribers throughout 2021, while linear affiliate revenues were also up year-over-year as rate increases continued to outpace subscriber declines. Our fully distributed subscribers were down 4% as were total portfolio subscribers when correcting for the impact of the sale of our Great American Country network in early June last year.

Yes, cords are still being cut, especially last year, but the story of cable for the last several years has been the jettisoning of cost-driven subscribers in favor of charging full price for people who actually want linear TV, which, in the long run, means that the linear TV bundle is primarily the sports and news bundle. Warner Bros. is well-placed for this new world, thanks to its combination of sports rights on TNT and TBS, along with CNN. That, though, isn’t a guarantee of profitability; Zaslav noted in an answer explaining why news was more profitable than sports:

When it comes to sports, we’re very careful about sports. And the TNT and Warner team was clever about getting long term rights which we’re going to get a lot of benefit from, but sports are rented and news is scalable.

The unscripted content that Discovery specializes in is even more scalable, and far cheaper; now, instead of negotiating with cable providers like Comcast for a collection of channels that customers like, but don’t necessarily need (particularly since Discovery+ is an option), Warner Bros. Discovery will be negotiating for a bundle that includes sports and news and filler. Those sports rights may eat up a lot of TNT and TBS’s carriage fees; the real money will be made on the extra pennies added to the rest of the channels in the portfolio.

And then, of course, that money can be spent on streaming: sure, Netflix has the advantage of having a larger customer base, but no one — other than Netflix executives until a month ago — said that you could only use subscription dollars and nothing else to acquire streaming content.

Advertising’s Continued Strength

Then there is advertising. First — and in contrast to Netflix’s agonizing on the matter — I enjoyed how matter-of-fact Zaslav was about having ad-supported streaming tiers:

In streaming, we have a massive opportunity to reach the widest possible addressable market by offering a range of tiers, all with the most compelling and complete portfolio of content. A premium and attractively priced ad-free direct-to-consumer product, a lower-priced ad-light tier, something we have had tremendous success with and is our highest ARPU product, and in some very price-sensitive markets outside the United States, we can even offer an advertiser-only product.

Secondly, Warner Bros. Discovery can provide a one-stop shop for advertisers across streaming and TV; Zaslav said in his opening remarks:

The combined strengths of both organizations’ client relationships, advanced advertising, programmatic, sponsorships and direct-to-consumer, ad-light streaming services, all positioned the company with a unique hand. I’ve personally spent quite a bit of time with key advertisers and agencies, and I’m so impressed with the combined capability of our platforms and our ability to uniquely serve the needs of our clients, including integrating sports alongside our broad entertainment offerings.

The bit about integrating sports refers to the fact that Warner Bros. actually had multiple advertising teams (one for sports and one for everything else — but none for streaming); I’m not entirely clear how the company ended up in this position, but given the fact the company’s businesses were built in an era where ad agencies sat in the middle between advertisers on one side and inventory on the other, it makes a certain sort of sense. Today, though, the goal is to be as large and as integrated as possible, the better to share data and provide effective targeting at scale; that offering is particularly compelling given that streaming is the best place to reach all of those people that cut the cord.

Still, even with the cord-cutting, TV advertising continues to be a good business; Zaslav reflected:

We recognize that 4% of subscribers are down and viewership on the platform is down…Long-term, there’s no question that the business is challenging, but CPMs are increasing, advertisers still are looking for inventory, because it’s the most effective inventory in long-form video. And look, remember, broadcast for a period of 20 years was declining and CPMs were increasing. I was at NBC in the mid-’90s when Welsh was saying this can’t continue. We can’t have smaller and smaller audiences and make more and more money. And I think he was right or maybe he will be right eventually, but it’s almost 30 years later and the advertisers are still paying more than the hurdle rate of decline.

So, we will be leaning in with efficiencies and effectiveness to our traditional business, which generates an awful lot of free cash flow…We now have the same or in many cases, the largest reach on television in the US. And the ability to use our own inventory to promote to and from all of our products and the efficiency of doing that and the cost savings of doing it, I think is a big plus for us.

There are a few things going on here, and both go back to an Article I wrote in 2016 called TV Advertising’s Surprising Strength — and Inevitable Fall. The key insight in that piece was that huge swathes of the economy, from large CPG companies to big box retailed to auto makers, were all built around TV advertising, which meant they would prop up the medium far longer than people thought:

Brands uniquely suited to TV are probably by definition less suited to digital advertising, which at least to date has worked much better for direct response marketing. No one is going to click a link in their feed to buy a car or laundry detergent, and a brick-and-mortar retailer doesn’t want to encourage shopping to someone already online. So after a bit of experimentation, they’re back with TV.

Still, I think Facebook and Snapchat in particular will figure brand advertising out: both have incredibly immersive advertising formats, and both are investing in ways to bring direct response-style tracking to brand advertising, including tracking visits to brick-and-mortar retailers.

Six years on, and “Surprising Strength” remains correct, while “Inevitable Fall” is the prediction that is looking shakier: to the extent that brand advertising is going digital, a lot of the shift seems to be doing so primarily as streaming video ads, a shift that will only accelerate as Netflix, HBO Max, and Disney all launch ad products. The most privacy-invasive practices of Facebook et al, meanwhile, have been curtailed (and rightly so — products that cross over into real-world tracking are where I have always drawn a hard line).

Just as important, though, are the impact of changes like ATT on the small and medium-sized businesses that were threatening the biggest advertisers like a hundred duck-sized horses: by destroying their ability to effectively coordinate their advertising spend, ATT and similar regulations have breathed new life into the old ecosystem, which ultimately plays to the benefit of the largest advertising sellers, including Warner Bros. Discovery.

Reasons for Optimism

Like I said, I’m pretty optimistic about Warner Bros. Discovery, which is why I appreciated the Wall Street Journal article I opened with: it’s fair to wonder if the exact sort of clarity of thinking and explicit commitment to financial results that Zaslav demonstrated on that earnings call will translate into managing talent and navigating Hollywood, particularly given the huge debt load the new company is carrying.

That noted, the other reason to be bullish is that Warner Bros. Discovery’s strategy is, in contrast to Netflix, back to the future; Zaslav said at the beginning of the call:

These last few months in our industry have been an important reminder that while technology will continue to empower consumers of video entertainment, the recipe for long-term success is still made up of a few key ingredients. Number one, world-class IP content that is loved all over the globe; two, distribution of that content on every platform and device where consumers want to engage, whether it’s theatrical or linear or streaming; three, a balanced monetization model that optimizes the value of what we create and drives diversified revenue streams; and four, finally, durable and sustainable free cash flow generation.

This isn’t anything new: the Hollywood model has always been about creating compelling content once and then monetizing it through every possible means; in Zaslav’s view a self-owned streaming platform is just an addition to the old model, not a wholesale replacement.

Moreover, things like movies in theaters still have their advantages: they build IP and build awareness for the other content models, above and beyond the money they make. Oh, and to take this update full circle, they also make talent like Eastwood happy. To that end, I suspect in the long run that flexibility and pragmatism in content distribution, combined with real discipline about cash flow, will prove to be more compelling than the same combination in reverse.

Cable’s Last Laugh

Posted on

If there is one industry people in tech are eternally certain is doomed, it is cable. However, the reality is that cable is both stronger than ever and poised for growth; the reasons why are instructive to not just tech industry observers, but to tech companies themselves.

The Creation of Cable

Robert J. Tarlton was 29 years old, married with a son, when he volunteered to fight in World War II; thanks to the fact he owned a shop in Lansford, Pennsylvania that sold radios, he ended up repairing them all across the European Theater, learning about not just reception but also transmission. After the war Tarlton re-opened his shop, when Motorola, one of his primary suppliers, came out with a new television.

Tarlton was intrigued, but he had a geography problem; the nearest television station was in Philadelphia, 71 miles away; in the middle lie the Pocono Mountains, and mountains aren’t good for reception:

Lansford is separated from Philadelphia by the Pocono Mountains

It turned out, though, that some people living in Summit Hill, the next community over, could get the Philadelphia broadcast signal; that’s where Tarlton sold his first television sets. Of course Tarlton couldn’t demonstrate this new-fangled contraption; his shop was in Lansford, not Summit Hill. However, it was close to Summit Hill: what if Tarlton could place an antenna further up the mountain in Summit Hill and run a cable to his shop? Tarlton explained in an interview in 1986:

Lansford is an elongated town. It’s about a mile and a half long and there are about eight parallel streets bisected with cross streets about every 500 feet. There are no curves; everything is all laid out in a nice symmetrical pattern. Our business place was about three streets from the edge of Summit Hill and Lansford sits on kind of a slope. The edge of Lansford inclines from about a thousand feet above sea level to about fifteen hundred feet above sea level in Summit Hill. So to get television into our store, my father and I put an antenna partly up the mountain. No, we didn’t go all the way up, but we put up an antenna, kind of a crude arrangement, and then from tree to tree we strung a twin lead that was used in those days as a transmission line. We ran this twin lead, crossed a few streets, and into our store. And we had television.

The basic twin lead was barely functional, but that didn’t stop everyone from demanding a television with a haphazard wire to their house; Tarlton realized that new coaxial cable amplifiers designed for a single property could be chained together, re-amplifying the signal so it could reach multiple properties. After getting all of the other electronics retailers on board — Tarlton knew that a clear signal would sell more TV sets, but that everyone needed to use the same system — the first commercial cable system was born, and it sold itself. Tarlton reflected:

You didn’t have to advertise. You had to keep your door locked because the people were clamoring for service. They wanted cable service. You certainly didn’t have to advertise.

People couldn’t get enough TV; Tarlton explained:

Cable is dependent upon advances in technology because people who originally saw one channel wanted to see the second channel, wanted to see the third, and after you had five, they wanted more. So it was a case of more begets more. At one time three channels seemed to be quite sufficient but when we added one more channel, it created a new interest in the cable system. People then had variety. They had alternatives. At one time later five channels seemed enough. As a matter of fact, a man who is often quoted, former FCC Commissioner Ken Cox, said that five channels was enough, and that’s quite a story in itself. The engineers were able to continually refine equipment to add more channels…All these technical advances‑‑continuing advances, automatic gain control, automatic temperature compensation, etc. have made cable what it is today.

One of the most important technological developments was satellite: now cable systems could get signals both more reliably and from far further away; this actually flipped geography on its head. Tarlton said:

At that time we went to the highest possible point to look to the transmitter. Today with satellites, we go to the lowest possible point because we don’t want the interference from other signals. So it is ironic that we have changed so much from what we used to do.

Within these snippets is everything that makes the cable business so compelling:

  • Cable is in high demand because it provides the means to get what customers most highly value.
  • Cable works best both technologically and financially when it has a geographic monopoly.
  • Cable creates demand for new supply; technological advances enable more supply, which creates more demand.

It’s that last bit about satellites being better on lower ground that stands out to me, though: as long as you control the wires into people’s houses you can and should be pragmatic about everything else.

Cable’s Evolution

Tarlton would go on to work for a company called Jerrold Electronics, which pivoted its entire business to create equipment for cities that wanted to emulate Lansford’s system; Tarlton would lead the installation of cable systems across the United States, which for the first two decades of cable mostly retransmitted broadcast television.

The aforementioned satellite, though, led to the creation of national TV stations, first HBO, and then WTCG, an independent television station in Atlanta, Georgia, owned by Ted Turner. Turner realized he could buy programming at local rates, but sell advertising at national rates via cable operators eager to feed their customers’ hunger for more stations. Turner soon launched a cable only channel devoted to nothing but news; he called it the Cable News Network — CNN for short (WTCG would later be renamed TBS).

Jerrold Electronics, meanwhile, spun off one of the cable systems it built in Tupelo, Mississippi to an entrepreneur named Ralph Roberts; Roberts proceeded to systematically buy up community cable systems across the country, moving the company’s headquarters to Philadelphia and renaming it to Comcast Corporation (Roberts would eventually hand the business off to his son, Brian). Consolidation in the provision of cable service proceeded in conjunction with consolidation in the production of content, an inevitable outcome of the virtuous cycle I noted above:

  • Cable companies acquired customers who wanted access to content
  • Studios created content that customers demanded

The more customers that a cable company served, the stronger their negotiating position with content providers; the more studios and types of content that a content provider controlled the stronger their negotiating position with cable providers. The end result were a few dominant cable providers (Comcast, Charter, Cox, Altice, Mediacom) and a few dominant content companies (Disney, Viacom, NBC Universal, Time Warner, Fox), tussling back-and-forth over a very profitable pie.

Then came Netflix, and tech industry crowing about cord cutting.

Netflix and other streaming services were obviously bad for television: they did the same job but in a completely different way, leveraging the Internet to provide content on-demand, unconstrained by the linear nature of television that was a relic of cable’s origin with broadcast TV. Here, though, cable’s ownership of the wires was an effective hedge: the same wires that delivered linear TV delivered packet-based Internet content.

Moreover, this didn’t simply mean that cable’s TV losses were made up for by Internet service: Internet service was much higher margin because companies like Comcast didn’t need to negotiate with a limited number of content providers; everything on the Internet was free. This has meant that the fortunes of cable companies has boomed over the last decade, even as cord-cutting has cut the cable TV business by about a third.

Cable companies today, though, are yet another category down from their pandemic highs, thanks to fear that the broadband growth story is mostly over; fiber offers better performance, 5G opens the door to wireless in the home, and anyone who doesn’t have broadband now is probably never going to get it. I think, though, this underrates the strategic positioning of cable companies, and ignores the industry’s demonstrated ability to adapt to new strategic environments.

The Wireless Opportunity

From the Wall Street Journal in 2011:

Verizon Wireless will pay $3.6 billion to buy wireless spectrum licenses from a group of cable-television companies, bringing an end to their years-long flirtation with setting up its own cellphone service. The sellers—Comcast Time Warner Cable Inc. and Bright House Networks—acquired the spectrum in a government auction in 2006 and now will turn it over to the country’s biggest wireless carrier at more than a 50% markup. While cable companies have dabbled with wireless, the spectrum has largely sat around unused, prompting years of speculation about the industry’s intentions…

Under the deal, Verizon Wireless will be able to sell its service in the cable companies’ stores. The carrier, in turn, will be able to sell the cable companies’ broadband, video and voice services in its stores. Verizon’s FiOS service only reaches 14% of U.S. households, according to Bernstein Research. In four years’ time, the cable companies will have the option to buy service on Verizon’s network on a wholesale basis and then resell it under their own brand.

The joint marketing arrangement “amounts to a partnership between formerly mortal enemies,” wrote Bernstein analyst Craig Moffett in a research note.

Moffett would later partner with Michael Nathanson to form their own independent research firm (called MoffettNathanson, natch); Moffett released an updated report last month that explained how that Verizon deal ended up playing out:

When Comcast and Time Warner Cable sold their AWS-1 spectrum to Verizon back in 2011, they believed at the time that they were walking away from ever becoming facilities-based wireless players. They therefore viewed it as imperative that the sale come with an MVNO agreement with Verizon to compensate for that forfeiture. They got precisely what they wanted, an MVNO contract with Verizon that was described at the time as “perpetual and irrevocable“…

It took another six years after that transaction before Comcast finally launched Xfinity Mobile in mid-2017. Charter [which merged with Time Warner Cable, acquiring the latter’s MVNO rights] followed suit a year later…in four short years, the Cable operators have become the fastest growing wireless providers in the country, accounting for nearly 30% of wireless industry net additions. Cable’s 7.7 million mobile lines represent ~2.5% of the U.S. mobile phone market (including prepaid and postpaid phones).

As Moffett notes, this growth is particularly impressive given that most cable companies couldn’t feasibly offer family plans under the terms of the original deal, which was negotiated before such a concept even existed; the deal has been re-negotiated, though, and almost certainly to the cable companies’ advantage: if Verizon is going to lose customers to an MVNO, it would surely prefer said MVNO be on their network; this means that the cable companies have negotiating leverage.

What, though, makes the cable companies such effective MVNOs? One of the most interesting parts of Moffett’s note was proprietary data about the amount of data used by cellular subscribers; it turns out that cable MVNO customers are far more likely to consume data over WiFi, perhaps because of cable company out-of-home WiFi hot spots. This could become even more favorable in the future as cable companies build out Citizens Broadband Radio Service (CBRS) service, particularly in dense areas where cable companies have wires from which to hang CBRS transmitters. Moffett writes:

In a perfect world, Cable will offload traffic onto their own facilities where doing so is cheap (high density, high use locations) and leave to Verizon the burden of carrying traffic where doing so is/would be expensive. Because the MVNO agreement is “perpetual and irrevocable,” and is based on average prices (i.e., the same price everywhere, whether easy or hard to reach), Cable is presented with a perfect ROI arbitrage; they can take the high ROI parts of the network for themselves, and leave the low ROI parts of the network to their MVNO host… all without sacrificing anything with respect to their national coverage footprint.

One can understand why Verizon gave the cable companies these rights in 2011; the phone company desperately needed spectrum, and besides, everyone knew that MVNOs could never be economically competitive with their wholesale providers, who were the ones that actually made the investments in the network.

That’s the thing, though: cable companies had their own massive build-out, one that was both much older in its origins and, because it connected the house, actually carried more data. This was to the benefit of Verizon and other cellular providers, of course: a Verizon iPhone uses WiFi at home just as much as a Comcast iPhone does; here, though, it was cable that had internet economics on its side: there is no competitive harm in giving equal access to an abundant resource; it is cellular access that is scarce, which means that the cable companies’ MVNO deal, in conjunction with their out-of-home Internet access options, gives them a meaningful advantage.

Customer Acquisition

Moffett spends less time on customer acquisition; anecdotally speaking, it’s clear that Charter’s Spectrum, which provides the cable service I consume (via the excellent Channels application), is pushing wireless service hard: phones are front-and-center in their stores, and Spectrum wireless commercials fill local inventory. Moreover, this is a well-trodden playbook: cable companies came to dominate the fixed line phone service business simply because it was easier to get your TV, Internet, and phone all in the same place (and of those, the most important was TV, and now Internet); it’s always easier to upsell an existing subscriber than it is to acquire a new one.

Of course cable companies long handled customer acquisition for content creators — that was the cable bundle in a nutshell. What is interesting is how this customer acquisition capability is attractive to the companies undoing that bundle: Netflix, for example, put its service on Comcast’s set-top boxes in 2016, and made a deal for Comcast to sell its service in 2018. I wrote in a Daily Update at the time:

Netflix, meanwhile, is laddering up again: the company doesn’t actually need a billing relationship with end users, it just needs ever more end users (along with the data about what they watch) to spread its fixed content costs more widely; the company said in its shareholder letter that:

These relationships allow our partners to attract more customers and to upsell existing subscribers to higher ARPU packages, while we benefit from more reach, awareness and often, less friction in the signup and payment process. We believe that the lower churn in these bundles offsets the lower Netflix ASP.

What is particularly interesting is that this arrangement moves the industry closer to the endgame I predicted in The Great Unbundling…

“Endgame” was a bit strong: that Article was, as the title says, about unbundling; one of my arguments was that the traditional cable TV bundle would become primarily anchored on live sports and news, while most scripted content went to streaming. That is very much the case today (TBS, for example, is abandoning scripted content, while becoming ever more reliant on sports). What was a mistake was insinuating that this was the “end”; after all, as Jim Barksdale famously observed, the next step after unbundling is bundling.

To that end, Netflix + Xfinity TV service was a bundle of sorts, but the real takeaway was that Comcast was fine with being simply an Internet provider (which ended up helping with TV margins, since cable companies mostly gave up on fighting to keep cord cutters). Shortly after the Netflix deal Comcast launched Xfinity Flex, a free 4K streaming box for Internet-only subscribers that included a storefront for buying streaming services (which would be billed by Comcast). You can even subscribe to digital MVPDs like YouTube TV!

The free Xfinity Flex streaming box

The first takeaway of an offering like Xfinity Flex ties into the wireless point: cable companies already have a billing relationship with the customer — because they provide the most essential utility for accessing what is most important to said customer — which makes them particularly effective at customer acquisition. That is why Netflix, YouTube, etc. are all willing to pay a revenue share for the help.

The Great Rebundling?

The second takeaway, though, is that the cable companies are better suited than almost anyone else to rebundle for real. Imagine a “streaming bundle” that includes Netflix, HBO Max, Disney+, Paramount+, Peacock, etc., available for a price that is less than the sum of its parts. Sounds too good to be true, right? After all, this kind of sounds like what Apple was envisioning for Apple TV (the app) before Netflix spoiled the fun; I wrote in Apple Should Buy Netflix in 2016:

Apple’s desire to be “the one place to access all of your television” implies the demotion of Netflix to just another content provider, right alongside its rival HBO and the far more desperate networks who lack any sort of customer relationship at all. It is directly counter to the strategy that has gotten Netflix this far — owning the customer relationship by delivering a superior customer experience — and while Apple may wish to pursue the same strategy, the company has no leverage to do so. Not only is the Apple TV just another black box that connects to your TV (that is also the most expensive), it also, conveniently for Netflix, has a (relatively) open app platform: Netflix can deliver their content on their terms on Apple’s hardware, and there isn’t much Apple can do about it.

Six years on and Netflix is in a much different place, not only struggling for new customers but also dealing with elevated churn. Owning the customer may be less important than simply having more customers, particularly if those customers are much less likely to churn. After all, that’s one of the advantages of a bundle: instead of your streaming service needing to produce compelling content every single month, you can work as a team to keep customers on board with the bundle.

The key point about a bundle, as longtime YouTube executive and Coda CEO Shishir Mehrotra has written, is that it minimizes SuperFan overlap while maximizing CasualFan overlap; in other words, effective bundles have more disparate content that you are vaguely interested in, instead of a relatively small amount of focused content that you care about intensely. This makes the bundle concept even more compelling to new entrants in the streaming wars, who may not have as large of libraries as Netflix, and certainly don’t have as many subscribers over which to spread their content costs.

Moreover, the fact that the streaming services have largely done their damage to traditional TV, leaving the cable TV bundle as the sports and news bundle, means it is actually viable to create a lower-priced bundle than what was previously available (if you don’t want sports). After all, you’re not cannibalizing TV, but rather bringing together what has long since been broken off (unbundling then bundling!).

This isn’t something that is going to happen overnight: despite the fact that bundles are good for everyone it is hard to get independents into a bundle as long as they are growing; Netflix’s recent struggles are encouraging in this regard, particularly if other streaming services start to face similar headwinds. Moreover, cable companies are not the only entities that will seek to pull something like this off: Apple, Amazon, Google, and Roku already make money from revenue shares on streaming subscriptions they sell; all of them sell devices that can be used as interfaces for selling a bundle. And, of course, there are more Internet providers than just the cable companies: there is fiber, wireless, and even Starlink.

The breadth of the cable company bundle, though, is unmatched: not only might it include streaming services, but also linear TV; more than that, this is the company selling you Internet access, and increasingly wireless phone service. That gives even more latitude for discounts, and perks like no data caps on streamed content, not just at home but also on your phone.

All of these advantages go back to Robert J. Tarlton and Lansford, Pennsylvania. A recurring point on Stratechery this past year has been the durability and long-term potential inherent in technology rooted in physical space. I wrote in Digital Advertising in 2022:

Real power in technology comes from rooting the digital in something physical: for Amazon that is its fulfillment centers and logistics on the e-commerce side, and its data centers on the cloud side. For Microsoft it is its data centers and its global sales organization and multi-year relationships with basically every enterprise on earth. For Apple it is the iPhone, and for Google is is Android and its mutually beneficial relationship with Apple (this is less secure than Android, but that is why Google is paying an estimated $15 billion annually — and growing — to keep its position). Facebook benefited tremendously from being just an app, but the freedom of movement that entailed meant taking a dependency on iOS and Android, and Apple has exploited that dependency in part, if not yet in full.

For cable companies, power comes from a wire; it would certainly be ironic if the cord-cutting trumpeted by tech resulted in cable having even more leverage over customers and their wallets than the pre-Internet era.

Beyond Aggregation: Amazon as a Service

Posted on

Five months and $134 billion in market cap ago (before the stock slipped by 68%), Bloomberg Businessweek purported to explain How Shopify Outfoxed Amazon to Become the Everywhere Store. One of the key parts of the story was how Shopify pulled one over on Amazon seven years ago:

An even more critical event came a few months after the IPO. Amazon also operated a service that let independent merchants run their websites, called Webstore. Bang & Olufsen, Fruit of the Loom, and Lacoste were among the 80,000 or so companies that used it to run their online shops. If he wanted to, Bezos surely had the resources and engineering prowess to crush Shopify and steal its momentum.

But Amazon execs from that time admit that the Webstore service wasn’t very good, and its sales were dwarfed by all the rich opportunities the company was seeing in its global marketplace, where customers shop on Amazon.com, not on merchant websites…In late 2015, in one of Bezos’ periodic purges of underachieving businesses, he agreed to close Webstore. Then, in a rare strategic mistake that’s likely to go down in the annals of corporate blunders, Amazon sent its customers to Shopify and proclaimed publicly that the Canadian company was its preferred partner for the Webstore diaspora. In exchange, Shopify agreed to offer Amazon Pay to its merchants and let them easily list their products on Amazon’s marketplace. Shopify also paid Amazon $1 million—a financial arrangement that’s never been previously reported.

Bezos and his colleagues believed that supporting small retailers and their online shops was never going to be a large, profitable business. They were wrong—small online retailers generated about $153 billion in sales in 2020, according to AMI Partners. “Shopify made us look like fools,” says the former Amazon executive.

If only we could all make such excellent mistakes; Amazon’s move looks like a strategic masterstroke.

Shopify’s Revenue Streams

Three major things have changed, will change, or should change about Shopify’s business in the years since the company made that deal with Amazon.

What has changed is the composition of Shopify’s business. While the company started out with a SaaS model, the business has transformed into a commission-based one:

Shopify's two revenue streams

“Subscription Solutions” are Shopify’s platform fees, including the cost to use the platform (or upgrade to the company’s Pro offering), commissions from the sales of themes and apps, and domain name registration. “Merchant Solutions”, meanwhile, are all of the fees that are generated from ongoing sales; the largest part of this are payment processing fees from Shopify Payments, but other fees include advertising revenue, referral fees, Shopify Shipping, etc.

It’s the shipping part that is line for big changes: while Shopify first announced the Shopify Fulfillment Network back in 2019, it is only recently that the company has committed to actually building out important pieces of said network on its own, the better to compete with Amazon’s full-scale offering.

As for what should change, I argued back in February that Shopify needed to build out an advertising network; this recommendation is more pertinent than ever, in large part because the second item on this list might be in big trouble.

Buy With Prime

From the Wall Street Journal:

Amazon.com Inc. is extending some of the offerings of its popular Prime membership program to merchants off its platform with a new service that embeds the online retailing giant’s payment and fulfillment options onto third-party sites. Called Buy with Prime, the service will allow merchants to show the Prime logo and offer Amazon’s speedy delivery options on products listed on their own websites…

The company said the Buy with Prime offer will be rolled out by invitation only through 2022 for those who already sell on Amazon and use the company’s fulfillment services. Later, Amazon plans to extend Buy with Prime to other merchants, including those that don’t sell on its platform. Participating merchants will use the Prime logo and display expected delivery dates on eligible products. Checkout will go through Amazon Pay and the company’s fulfillment network. Amazon will also manage free returns for eligible orders.

This is a move that you could see coming for a long time; back in 2016 I wrote an article called The Amazon Tax that explained that the best way to understand Amazon as a whole was to understand Amazon Web Services (AWS):

The “primitives” model modularized Amazon’s infrastructure, effectively transforming raw data center components into storage, computing, databases, etc. which could be used on an ad-hoc basis not only by Amazon’s internal teams but also outside developers:

A drawing of The AWS Layer

This AWS layer in the middle has several key characteristics:

  • AWS has massive fixed costs but benefits tremendously from economies of scale
  • The cost to build AWS was justified because the first and best customer is Amazon’s e-commerce business
  • AWS’s focus on “primitives” meant it could be sold as-is to developers beyond Amazon, increasing the returns to scale and, by extension, deepening AWS’ moat

This last point was a win-win: developers would have access to enterprise-level computing resources with zero up-front investment; Amazon, meanwhile, would get that much more scale for a set of products for which they would be the first and best customer.

As I noted in that article, the AWS model was being increasingly applied to e-commerce as Amazon shifted from being a retailer to being a services provider:

Prime is a super experience with superior prices and superior selection, and it too feeds into a scale play. The result is a business that looks like this:

A drawing of The Transformation of Amazon’s E-Commerce Business

That is, of course, the same structure as AWS — and it shares similar characteristics:

  • E-commerce distribution has massive fixed costs but benefits tremendously from economies of scale
  • The cost to build-out Amazon’s fulfillment centers was justified because the first and best customer is Amazon’s e-commerce business
  • That last bullet point may seem odd, but in fact 40% of Amazon’s sales (on a unit basis) are sold by 3rd-party merchants; most of these merchants leverage Fulfilled-by-Amazon, which means their goods are stored in Amazon’s fulfillment centers and covered by Prime. This increases the return to scale for Amazon’s fulfillment centers, increases the value of Prime, and deepens Amazon’s moat

My prediction in that Article was that Amazon’s burgeoning logistics business would eventually follow the same path:

It seems increasingly clear that Amazon intends to repeat the model when it comes to logistics…how might this play out? Well, start with the fact that Amazon itself would be this logistics network’s first-and-best customer, just as was the case with AWS. This justifies the massive expenditure necessary to build out a logistics network that competes with UPS, Fedex, et al, and most outlets are framing these moves as a way for Amazon to rein in shipping costs and improve reliability, especially around the holidays.

However, I think it is a mistake to think that Amazon will stop there: just as they have with AWS and e-commerce distribution I expect the company to offer its logistics network to third parties, which will increase the returns to scale, and, by extension, deepen Amazon’s eventual moat

Today Amazon’s logistics is massive and fully integrated from the fulfillment center to the doorstep, even though it only serves Amazon; the obvious next step is opening it up to non-Amazon retailers, and that is exactly what is happening.

Beyond Aggregation

At first glance, this might seem like a bit of a surprise; after all, Stratechery is well known for describing Aggregation Theory, which is predicated on controlling demand. In the case of Amazon that has meant controlling the website where customers order goods — Amazon.com — even if those goods were sold by 3rd-party merchants. Why would Amazon give that up?

The reasoning is straightforward: while Amazon has had Aggregator characteristics, the company’s business model and differentiation has always been rooted in the real world, which, by extension, means it is not an Aggregator at all. I noted in 2017’s Defining Aggregators that Aggregators benefit from zero marginal costs, which only describes a certain set of digital businesses like Google and Facebook:

Companies traditionally have had to incur (up to) three types of marginal costs when it comes to serving users/customers directly.

  • The cost of goods sold (COGS), that is, the cost of producing an item or providing a service
  • Distribution costs, that is the cost of getting an item to the customer (usually via retail) or facilitating the provision of a service (usually via real estate)
  • Transaction costs, that is the cost of executing a transaction for a good or service, providing customer service, etc.

Aggregators incur none of these costs:

  • The goods “sold” by an Aggregator are digital and thus have zero marginal costs (they may, of course, have significant fixed costs)
  • These digital goods are delivered via the Internet, which results in zero distribution costs
  • Transactions are handled automatically through automatic account management, credit card payments, etc.

This characteristic means that businesses like Apple hardware and Amazon’s traditional retail operations are not Aggregators; both bear significant costs in serving the marginal customer (and, in the case of Amazon in particular, have achieved such scale that the service’s relative cost of distribution is actually a moat).

Amazon’s control of demand has been — and will continue to be — a tremendous advantage; Amazon not only has power over its suppliers, but it also gets all of the relevant data from consumers, which it can feed into a self-contained ad platform that is untouched by regulation from either governments or Apple.

At the same time, limiting a business to customer touchpoints that you control means limiting your overall addressable market. This may not matter in markets where there are network effects (which means you appeal to everyone) and you are an Aggregator dealing with zero marginal costs (and thus can realistically cover every consumer); in the case of e-commerce, though, Amazon will never be the only option, particularly given The Anti-Amazon Alliance working hard to reach consumers.

A core part of the Anti-Amazon Alliance are companies that have invested in brands that attract customers on their own; these companies don’t need to be on Amazon fighting to be the answer to generic search terms, but can rather drive customers to their own Shopify-powered websites both organically and via paid advertising (Facebook is a huge player in the Anti-Amazon Alliance). Still, every customer that visits these websites has an Amazon-driven expectation in terms of shipping; Shippo CEO Laura Behrens Wu told me in a Stratechery interview:

Consumers have those expectations from Amazon that shipping should be free, it should be two days, and whatever those are expectations are, returns should be free. That is still carried over when I’m buying on this branded website. If the expectations are not met, consumers decide to buy somewhere else…Merchants are constantly trying to play catch up, whatever Amazon is doing they need to follow suit.

Now Amazon has — or soon will have, in the case of Shopify-only merchants — a solution: the best way to get an Amazon-like shipping experience is to ship via Amazon. And, in contrast to the crappy Webstore product, you can keep using Shopify and its ecosystem for your website. Amazon may have given away business to Shopify in 2015, but that doesn’t much matter if said business ends up being a commoditized complement to Amazon’s true differentiation in logistics. That business, thanks to the sheer expense necessary to build it out, has a nearly impregnable moat that is not only attractive to all of the businesses competing to be consumer touchpoints — thus increasing Amazon’s addressable market — but is also one that sees its moat deepen the larger it becomes.

Shopify’s Predicament and Amazon’s Opportunity

The reason this announcement is so damaging to Shopify goes back to the transformation in the company’s revenue I charted above: Amazon’s shipping solution requires the customer to pay with Amazon; I love Shop Pay’s checkout process, but it’s not as if I don’t have an Amazon account with all of my relevant details already included, and I absolutely trust Amazon’s shipping more than I do whatever option some Shopify merchant offers me. If I had a choice I’m taking the Amazon option every time.

Granted, I may not be representative; I’m a bit of a special case given where I live. What is worth noting, though, is that every transaction that Amazon processes is one not processed by Shopify, which again, is the company’s primary revenue driver. Moreover, the more volume that Amazon processes, the more difficult it will be for Shopify to get their own shipping solution to scale. This endangers the company’s current major initiative.

That is why I think the company needs to think more deeply than ever about advertising: the implication of Amazon being willing to be a services provider and not necessarily an Aggregator is that Amazon is surrendering some number of customer touchpoints to competitors; to put it another way, Amazon is actually making competitive customer touchpoints better — touchpoints that Shopify controls. Shopify ought to leverage that control.

That noted, there is a potential wrench in this plan: if Shopify doesn’t own payment processing, will they have sufficient data to build a competitive conversion-data-driven advertising product? Amazon — and Apple — would likely argue that that data is Amazon’s, but the merchant will obviously know what is going on (by the same token, will Amazon be able to tie this off-platform conversion data back to its own advertising product? It’s not clear what would stop them).

Shopify has two additional saving graces:

  • First, the fact that Amazon will be able to collect data is a big reason for many merchants not to use Amazon’s new offering. Shopify’s offering will always be differentiated in this regard.
  • Second, as the announcement noted, this is going to take Amazon a year or two to fully roll out, and lots of stuff can change in the meantime.

Moreover, while AWS has always been excellent at serving external customers — which it did before Amazon.com ever actually moved over — Amazon’s off-Amazon.com merchant offerings have never been particularly successful (including the aforementioned Webstores).

With that in mind, I think it’s meaningful that “Buy With Prime” is the first major initiative of new CEO Andy Jassy’s regime; I don’t think it’s an accident that it is so clearly inspired by AWS. AWS’s strength is its focus on infrastructure at scale; successfully moving e-commerce beyond aggregation to the same type of service business model would put his stamp on the company in a meaningful way, and, contra that quote in Bloomberg Businessweek, mark Jassy as nobody’s fool.

Back to the Future of Twitter

Posted on

Elon Musk wrote in a letter to Twitter’s board:

I invested in Twitter as I believe in its potential to be the platform for free speech around the globe, and I believe free speech is a societal imperative for a functioning democracy.

However, since making my investment I now realize the company will neither thrive nor serve this societal imperative in its current form. Twitter needs to be transformed as a private company.

As a result, I am offering to buy 100% of Twitter for $54.20 per share in cash, a 54% premium over the day before I began investing in Twitter and a 38% premium over the day before my investment was publicly announced. My offer is my best and final offer and if it is not accepted, I would need to reconsider my position as a shareholder.

Twitter has extraordinary potential. I will unlock it.

The vast majority of commentary about the Musk-Twitter saga has focused on the first three paragraphs: what does Musk mean by making Twitter more free speech oriented? Why doesn’t Musk believe he can work with the current board and management? Does Musk have the cash available to buy Twitter, and would the Twitter board accept his offer (no on the latter, but more on this below)?

The most interesting question of all, though, is the last paragraph: what potential does Musk see, and could he unlock it? For my part, not only do I agree the potential is vast, but I do think Musk could unlock it — and that itself has implications for the preceding paragraphs.

What is Twitter?

It’s popular on Twitter to point to a funny tweet or exchange and marvel that “This website is free“, or alternatively, “This app is free“. That raises the question, though, what is Twitter: is it a website or an app, or something different?

The answer, of course, is “All of the above”, but it’s worth being clear about the different pieces that make up Twitter; “Jin” made this illustration on Medium:

Twitter's architecture

Twitter is actually a host of microservices, including a user service (for listing a user’s timeline), a graph service (for tracking your network), a posting service (for posting new tweets), a profile service (for user profiles), a timeline service (for presenting your timeline), etc.; the architecture to tie all of these together and operate at scale all around the world is suitably complex.

The key thing to note, though, is that only the green boxes in the diagram above are actually user-facing; a dramatically simplified version of Twitter, that condenses all of those internal services to a big blue “Twitter” box and focuses on the green interfaces might look something like this:

A dramatic over-simplification of Twitter's architecture

Again, this is a dramatic oversimplification, but the important takeaway is that the user-facing parts of Twitter are distinct from — and, frankly, not very pertinent to — the core Twitter service.

Twitter’s API Drama

The general idea behind a services architecture is that various functionalities are exposed via application programming interfaces, more commonly known as APIs; a “client” will leverage these APIs to build an end user experience. There is no requirement that these clients be owned or managed by the centralized service, and for the first several years of Twitter’s existence, that is exactly how the service operated: Twitter the company ran the service and the Twitter.com website, while third-parties built clients that let you access Twitter first on the desktop and then on smartphones.

Mobile was an absolute boon for Twitter: the public messaging service, modeled on SMS, was a natural fit for a smartphone screen, and the immediacy of Twitter updates was perfectly suited to a device that was always connected to the Internet. The explosion in mobile usage, though, led to a situation where Twitter didn’t actually control the user experience for a huge portion of its users. This actually led to a ton of innovation: Twitterrific, for example, the earliest third party client, came up with the Twitter bird, the term “tweet” for a Twitter message, and early paradigms around replies and conversations. It also led to problems, the solutions to which fundamentally changed Twitter’s potential as a business.

The first problem that came from Twitter the service relying on third party clients is that the company, which descended into politics and backstabbing from the board level on down almost immediately, was drifting along without a business model; the obvious candidate was advertising, but the easiest way to implement advertising was to control the user interface (and thus insert ads — ads, including promoted tweets, are another distinct service from Twitter itself). Thus Twitter bought Tweetie, widely regarded as the best Twitter mobile client (I was a user), in April 2010, and rebranded and relaunched it as the official Twitter for iPhone app a month later.

The second problem is that starting in 2010, a Silicon Valley entrepreneur named Bill Gross (who invented search advertising) started trying to build his own Twitter monetization product called TweetUp; when Twitter acquired Tweetie and made clear they were going to monetize it via advertising, Gross started buying up multiple other third party Twitter clients with the idea of creating a competing network of clients that would monetize independently. Twitter responded in the short term by kicking several of Gross’s clients off of the platform for dubious terms-of-service violations, and in the long term by killing the 3rd party API for everyone. Clients could keep the users they had but could only add 100,000 more users — ever.

The net result of these two decisions was that Twitter, its architecture notwithstanding, would be a unified entity where Twitter the company controlled every aspect of the experience, and that that experience would be monetized via advertising.

Twitter’s Reality

Twitter has, over 19 different funding rounds (including pre-IPO, IPO, and post-IPO), raised $4.4 billion in funding; meanwhile the company has lost a cumulative $861 million in its lifetime as a public company (i.e. excluding pre-IPO losses). During that time the company has held 33 earnings calls; the company reported a profit in only 14 of them.

Given this financial performance it is kind of amazing that the company was valued at $30 billion the day before Musk’s investment was revealed; such is the value of Twitter’s social graph and its cultural impact: despite there being no evidence that Twitter can even be sustainably profitable, much less return billions of dollars to shareholders, hope springs eternal that the company is on the verge of unlocking its potential. At the same time, these three factors — Twitter’s financials, its social graph, and its cultural impact — get at why Musk’s offer to take Twitter private is so intriguing.

Start with the financials: Twitter’s business stinks. Yes, you can make an argument that this is due to mismanagement and poor execution — who enjoys seeing a stale promoted tweet about something that happened weeks ago?1 — but I have also made the argument that Twitter just isn’t well suited to direct response advertising in particular:

Think about the contrast between Twitter and Instagram; both are unique amongst social networks in that they follow a broadcast model: tweets on Twitter and photos on Instagram are public by default, and anyone can follow anyone. The default medium, though, is fundamentally different: Twitter has photos and videos, but the heart of the service is text (and links). Instagram, on the other hand, is nothing but photos and video (and link in bio).

The implications of this are vast. Sure, you may follow your friends on both, but on Twitter you will also follow news breakers, analysts, insightful anons, joke tellers, and shit posters. The goal is to mainline information, and Twitter’s speed and information density are unparalleled by anything in the world. On Instagram, though, you might follow brands and influencers, and your chief interaction with your friends is stories about their Turkey Day exploits. It’s about aspiration, not information, and the former makes a lot more sense for effective advertising.

It’s more than just the medium though; it’s about the user’s mental state as well. Instagram is leisurely and an escape, something you do when you’re procrastinating; Twitter is intense and combative, and far more likely to be tied to something happening in the physical world, whether that be watching sports or politics or doing work:

Instagram is a lean-back experience; on Twitter you lean forward

This matters for advertising, particularly advertising that depends on a direct response: when you are leaning back and relaxed why not click through to that Shopify site to buy that knick-knack you didn’t even know you needed, or try out that mobile game? When you are leaning forward, though, you don’t have either the time or the inclination.

That article made the argument for Twitter to move towards more of a subscription offering; that may be the wrong idea, but the bigger takeaway is that what Twitter has been trying to build for years just isn’t working, and the challenges aren’t just bad management. To put it another way, when it comes to Twitter’s business, there really isn’t much to lose, but Twitter could only risk losing what there is if it were a private company, free from the glare of public markets who, for very justifiable reasons, give Twitter’s management a very short leash.

What is valuable is that social graph: while Facebook understands who you know, Twitter, more than any other company, understands what its users are interested in. That is, in theory, much more valuable; said value is diminished by the fact that Twitter just doesn’t have that many users, relatively speaking; the users it has, though, are extremely influential, particularly given the important of Twitter in media, tech, and finance. For this group Twitter is completely irreplaceable: there is no other medium with a similar density of information or interest-driven network effects.

This, by extension, drives Twitter’s cultural impact: no, most people don’t get their news off of Twitter; the places they get their news, though, are driven by Twitter. Moreover, Twitter not only sets the agenda for media organizations, it also harmonizes coverage, thanks to a dynamic where writers, unmoored from geographic constraints or underlying business realities of their publications, end up writing for other writers on Twitter, oftentimes radicalizing each other in plain sight of their readership. Twitter itself is part of this harmonization, going so far as to censor politically impactful stories in the weeks before an election; it’s no surprise that when Musk says he wants to impose a stronger free speech ethos that the reaction is fierce and littered with motte-and-baileys (“actually we just care about limiting abuse and spam!”).

Back to the Future

This is all build-up to my proposal for what Musk — or any other bidder for Twitter, for that matter — ought to do with a newly private Twitter.

  • First, Twitter’s current fully integrated model is a financial failure.
  • Second, Twitter’s social graph is extremely valuable.
  • Third, Twitter’s cultural impact is very large, and very controversial.

Given this, Musk (who I will use as a stand-in for any future CEO of Twitter) should start by splitting Twitter into two companies.

  • One company would be the core Twitter service, including the social graph.
  • The other company would be all of the Twitter apps and the advertising business.

TwitterAppCo would contract with TwitterServiceCo to continue to receive access to the Twitter service and social graph; currently Twitter earns around $13/user/year in advertising, so you could imagine a price of say $7.50/user/year, or perhaps $0.75/user/month. TwitterAppCo would be free to pursue the same business model and moderation policies that Twitter is pursuing today (I can imagine Musk sticking with TwitterServiceCo, and the employees upset about said control being a part of TwitterAppCo).

However, that relationship would not be exclusive: TwitterServiceCo would open up its API to any other company that might be interested in building their own client experience; each company would:

  • Pay for the right to get access to the Twitter service and social graph.
  • Monetize in whatever way they see fit (i.e. they could pursue a subscription model).
  • Implement their own moderation policy.

This last point would cut a whole host of Gordian Knots:

  • Market competition would settle the question about whether or not stringent moderation is an important factor in success; some client experiences would be heavily moderated, and some wouldn’t be moderated at all.
  • The fact that everyone gets access to the same Twitter service and social graph solves the cold start problem for alternative networks; the reason why Twitter alternatives always fail is because Twitter’s network effect is so important.
  • TwitterServiceCo could wash its hands of difficult moderation decisions or tricky cultural issues; the U.S. might have a whole host of Twitter client options, while Europe might be more stringent, and India more stringent still. Heck, this model could even accommodate a highly-censored China client (although this is highly unlikely).

I strongly suspect that a dramatic increase in competition amongst Twitter client services would benefit TwitterServiceCo, growing its market in a way that hasn’t happened in years. What is most exciting, though, is the potential development of new kinds of services that don’t look like Twitter at all.

Step back a moment and think about the fundamental infrastructure of the Internet: we have a media protocol in HTTP/web, and a communications protocol in SMTP/email; what is missing is a notifications protocol. And yet, at the same time, if there is one lesson from mobile, it is just how important notifications are; a secondary consideration is how important identity is. If you can know how to reach someone, and have the means to do so, you are set, whether you be a critical service, an advertiser, or anything in-between. Twitter has the potential to fill that role: the ability to route short messages to a knowable endpoint accessible via a centralized directory has far more utility than political signaling and infighting. And yet, thanks to Twitter’s early decisions and lack of leadership, the latter is all the service is good for; no wonder user growth and financial results have stagnated!

A truly open TwitterServiceCo has the potential to be a new protocol for the Internet — the notifications and identity protocol; unlike every other protocol, though, this one would be owned by a private company. That would be insanely valuable, but it is a value that will never be realized as long as Twitter is a public company led by a weak CEO and ineffective board driving an integrated business predicated on a business model that doesn’t work.

Twitter’s Reluctance

The surest evidence of the Twitter board’s lack of imagination and ineffectiveness is that their response to Musk’s proposal is to further dilute existing shareholders as a means of denying Musk control. This is, in my estimation, clearly against the interest of Twitter shareholders (which, for what it’s worth, don’t in any meaningful way include Twitter’s board members); given Twitter’s performance over the last decade, though, this isn’t really a surprise.

Indeed, when you consider the fact that Twitter’s board members not only don’t own much of Twitter, but famously, barely use Twitter at all, it is easy to wonder if the actual goal is not financial results but rather harnessing that immense cultural impact. This suspicion only intensifies when you consider that the bidder in this case is one of the most successful entrepreneurs of all time: if there was one person in the world who could realize Twitter’s latent value, wouldn’t Musk be at the top of the list? And yet he is anathema, not for his business acumen, but despite it.

This, more than anything, makes me even more sure that my proposal for competition amongst Twitter client companies is essential: not only do I think that more competition would lead to dramatically more innovation, but it would also solve the problem of who decides what we see by undoing the centralization of that power and subjecting decisions to market forces. That this is unacceptable to some says more about their ultimate motivations than anything else.

I wrote a follow-up to this Article in this Daily Update.

  1. A friend sent me this promoted tweet on April 15, almost a full month since Kentucky had been eliminated from March Madness, and well over a week after the entire tournament had ended ↩

DALL-E, the Metaverse, and Zero Marginal Content

Posted on

Last week OpenAI released DALL-E 2, which produces (or edits) images based on textual prompts; this Twitter thread from @BecomingCritter has a whole host of example output, including Teddy bears working on new AI research on the moon in the 1980s:

Teddy bears working on new AI research on the moon in the 1980s

A photo of a quaint flower shop storefront with a pastel green and clean white facade and open door and big window:

A photo of a quaint flower shop storefront with a pastel green and clean white facade and open door and big window

And, in the most on-the-nose example possible, A human basking in the sun of AGI utopia:

A human basking in the sun of AGI utopia]

OpenAI has a video describing DALL-E on its website:

While the video does mention a couple of DALL-E’s shortcomings, it is quite upbeat about the possibilities; some excerpts:

Dall-E 2 is a new AI system from OpenAI that can take simple text descriptions like “A koala dunking a basketball” and turn them into photorealistic images that have never existed before. DALL-E 2 can also realistically edit and re-touch photos…

DALL-E was created by training a neural network on images and their text descriptions. Through deep learning it not only understands individual objects like koala bears and motorcycles, but learns from relationships between objects, and when you ask DALL-E for an image of a “koala bear riding a motorcycle”, it knows how to create that or anything else with a relationship to another object or action.

The DALL-E research has three main outcomes: first, it can help people express themselves visually in ways they may not have been able to before. Second, an AI-generated image can tell us a lot about whether the system understands us, or is just repeating what it’s been taught. Third, DALL-E helps humans understand how AI systems see and understand our world. This is a critical part of developing AI that’s useful and safe…

What’s exciting about the approach used to train DALL-E is that it can take what it learned from a variety of other labeled images and then apply it to a new image. Given a picture of a monkey, DALL-E can infer what it would look like doing something it has never done before, like paying its taxes while wearing a funny hat. DALL-E is an example of how imaginative humans and clever systems can work together to make new things, amplifying our creative potential.

That last line may raise some eyebrows: at first glance DALL-E looks poised to compete with artists and illustrators; there is another point of view, though, where DALL-E points towards a major missing piece in a metaverse future.

Games and Medium Evolution

Games have long been on the forefront of technological development, and that is certainly the case in terms of medium. The first computer games were little more than text:

A screenshot from Oregon Trail

Images followed, usually of the bitmap variety; I remember playing a lot of “Where in the world is Carmen San Diego” at the library:

A screenshot from "Where in the world is Carmen San Diego"

Soon games included motion as you navigated a sprite through a 2D world; 3D followed, and most of the last 25 years has been about making 3D games ever more realistic. Nearly all of those games, though, are 3D images on 2D screens; virtual reality offers the illusion of being inside the game itself.

Still, this evolution has had challenges: creating ever more realistic 3D games means creating ever more realistic image textures to decorate all of those polygons; this problem is only magnified in virtual reality. This is one of the reasons even open-world games are ultimately limited in scope, and gameplay is largely deterministic: it is through knowing where you are going, and all of your options to get there, that developers can create all of the assets necessary to deliver an immersive experience.

That’s not to say that games can’t have random elements, above and beyond roguelike games that are procedurally generated: the most obvious way to deliver an element of unpredictability is for humans to play each other, albeit in well-defined and controlled environments.

Social and User-Generated Content

Social networking has undergone a similar medium evolution as games, with a two-decade delay. The earliest forms of social networking on the web were text-based bulletin boards and USENET groups; then came widespread e-mail, AOL chatrooms, and forums. Facebook arrived on the scene in the mid-2000s; one of the things that helped it explode in popularity was the addition of images. Instagram was an image-only social network that soon added video, which is all that TikTok is. And, over the last couple of years in particular, video conferencing through apps like Zoom or Facetime have delivered 3D images on 2D screens.

Still, medium has always mattered less for social networking, just because the social part of it was so inherently interesting. Humans like communicating with other humans, even if that requires dialing up a random BBS to download messages, composing a reply, and dialing back in to send it. Games may be mostly deterministic, but humans are full of surprises.

Moreover, this means that social networking is much cheaper: instead of the platform having to generate all of the content, users generate all of the content themselves. This makes it harder to get a new platform off of the ground, because you need users to attract users, but it also makes said platform far stickier than any game (or, to put it another way, the stickiest games have a network effect of their own).

Feeds and Algorithms

The first iterations of social networking had no particular algorithmic component other than time: newer posts were at the top (or bottom). That changed with Facebook’s introduction of the News Feed in 2006. Now instead of visiting all of your friends’ pages you could simply browse the feed, which from the very beginning made decisions about what content to include, and in what order.

Over time the News Feed evolved from a relatively straightforward algorithm to one driven by machine learning, with results so inscrutable that it took Facebook six months to fix a recent rankings bug. The impact has been massive: not just Facebook but also Instagram saw huge increases in engagement and increased growth the better their algorithmically-driven feeds became; it was also great for monetization, as the same sort of signals that decided what content you saw also influenced what ads you were presented.

However, the reason why this discussion of algorithmically-driven feeds is in a different section than social networking is because the ultimate example of their power isn’t a social network at all: it’s TikTok. TikTok, of course, is all user-generated content, but the crucial distinction from Facebook is that you aren’t limited to content from your network: TikTok pulls in the videos it thinks you specifically are most interested in from across its entire network. I explained why this was a blindspot for Facebook in 2020:

What is interesting to point out is why it was inevitable that Facebook missed this: first, Facebook views itself first-and-foremost as a social network, so it is disinclined to see that as a liability. Second, that view was reinforced by the way in which Facebook took on Snapchat. The point of The Audacity of Copying Well is that Facebook leveraged Instagram’s social network to halt Snapchat’s growth, which only reinforced that the network was Facebook’s greatest asset, making the TikTok blindspot even larger.

TikTok combines the zero cost nature of user-generated content with a purely algorithmic feed that is divorced from your network; there is a network effect, in that TikTok needs lots of content to choose from, but it doesn’t need your specific network.

The Machine Learning Metaverse

I get that metaverses were so 2021, but it strikes me that the examples from science fiction, including Snow Crash and Ready Player One, were very game-like in their implementation. Their virtual worlds were created by visionary corporations or, in the case of the latter, a visionary developer who also included a deterministic game for ultimate ownership of the virtual world. Yes, third parties could and did build experiences with strong social components, most famously Da5id’s Black Sun club in Snow Crash, but the core mechanic — and the core economics — were closer to a multi-player game than anything else.

That, though, is exceptionally challenging in the real world: remember, creating games, particularly their art, is expensive, and the expense increases the more immersive the experience is. Social media, on the other hand, is cheap because it uses user-generated content, but that content is generally stuck on more basic mediums — text, pictures, and only recently video. Of course that content doesn’t necessarily need to be limited to your network — an algorithm can deliver anything on the network to any user.

What is fascinating about DALL-E is that it points to a future where these three trends can be combined. DALL-E, at the end of the day, is ultimately a product of human-generated content, just like its GPT-3 cousin. The latter, of course, is about text, while DALL-E is about images. Notice, though, that progression from text to images; it follows that machine learning-generated video is next. This will likely take several years, of course; video is a much more difficult problem, and responsive 3D environments more difficult yet, but this is a path the industry has trod before:

  • Game developers pushed the limits on text, then images, then video, then 3D
  • Social media drives content creation costs to zero first on text, then images, then video
  • Machine learning models can now create text and images for zero marginal cost

In the very long run this points to a metaverse vision that is much less deterministic than your typical video game, yet much richer than what is generated on social media. Imagine environments that are not drawn by artists but rather created by AI: this not only increases the possibilities, but crucially, decreases the costs.

Zero Marginal Content

There is another way to think about DALL-E and GPT and similar machine learning models, and it goes back to my longstanding contention that the Internet is a transformational technology matched only by the printing press. What made the latter revolutionary was that it drastically reduced the marginal cost of consumption; from The Internet and the Third Estate:

Meanwhile, the economics of printing books was fundamentally different from the economics of copying by hand. The latter was purely an operational expense: output was strictly determined by the input of labor. The former, though, was mostly a capital expense: first, to construct the printing press, and second, to set the type for a book. The best way to pay for these significant up-front expenses was to produce as many copies of a particular book that could be sold.

How, then, to maximize the number of copies that could be sold? The answer was to print using the most widely used dialect of a particular language, which in turn incentivized people to adopt that dialect, standardizing languages across Europe. That, by extension, deepened the affinities between city-states with shared languages, particularly over decades as a shared culture developed around books and later newspapers. This consolidation occurred at varying rates — England and France several hundred years before Germany and Italy — but in nearly every case the First Estate became not the clergy of the Catholic Church but a national monarch, even as the monarch gave up power to a new kind of meritocratic nobility epitomized by Burke.

The Internet has had two effects: the first is to bring the marginal cost of consumption down to zero. Even with the printing press you still needed to print a physical object and distribute it, and that costs money; meanwhile it costs effectively nothing to send this post to anyone in the world who is interested. This has completely upended the publishing industry and destroyed the power of gatekeepers.

The other impact, though, has been on the production side; I wrote about TikTok in Mistakes and Memes:

That phrase, “Facebook is compelling for the content it surfaces, regardless of who surfaces it”, is oh-so-close to describing TikTok; the error is that the latter is compelling for the content it surfaces, regardless of who creates it…To put it another way, I was too focused on demand — the key to Aggregation Theory — and didn’t think deeply enough about the evolution of supply. User-generated content didn’t have to be simply pictures of pets and political rants from people in one’s network; it could be the foundation of a new kind of network, where the payoff from Metcalfe’s Law is not the number of connections available to any one node, but rather the number of inputs into a customized feed.

Machine learning generated content is just the next step beyond TikTok: instead of pulling content from anywhere on the network, GPT and DALL-E and other similar models generate new content from content, at zero marginal cost. This is how the economics of the metaverse will ultimately make sense: virtual worlds needs virtual content created at virtually zero cost, fully customizable to the individual.

Of course there are many other issues raised by DALL-E, many of them philosophical in nature; there has already been a lot of discussion of that over the last week, and there should be a lot more. Still, the economic implications matter as well, and after last week’s announcement the future of the Internet is closer, and weirder, than ever.

Why Netflix Should Sell Ads

Posted on

The Information reported over the weekend that Netflix executives have told employees to keep an eye on the bottom line:

In two separate meetings over the past few weeks, Netflix executives cautioned employees to be more mindful about spending and hiring, according to three people familiar with the discussions. The comments, made at an employee town hall on Monday as well as during a management offsite held last month in Anaheim, Calif., come as the streaming giant grapples with sharply slowing subscriber growth…

Netflix has also been pondering steps that could help offset the revenue impact of the subscriber slowdown, including cracking down on people sharing the passwords to their accounts. While Netflix has long allowed such password sharing, it has become more common in the U.S. and other parts of the world than executives anticipated, the people said. This effort has been underway for about a year, however, well before the slowdown became apparent.

These are presented as two different issues, but there is a connection between them: Netflix should be hiring more people — a lot of them — and those people should be building a product that increases subscriber numbers and revenue. That product is advertising.

Netflix’s Business Model: Subscriptions

Netflix is, incredibly enough, 24 years old, and a subscription model has served the company well. Not that Netflix had much choice when it started: the company briefly sold DVDs online, before focusing exclusively on renting them; neither approach offered much surface area for advertising, and besides, the subscription model was revolutionary in its own right.

DVDs-by-mail was, from a certain perspective, inconvenient: you couldn’t simply drive to your local Blockbuster and peruse the selection; on the other hand, Netflix’s model gave you access to nearly every movie ever released, not just those in stock at your local store. The real innovation, though, was that business model: instead of paying to rent a DVD and being gouged with late fees, you could pay a set amount each month and keep the DVDs Netflix mailed to you as long as you wanted; send one back to get the next one in your queue.

Consumers loved it, and Netflix has stuck with the model even as the shift to streaming flipped their value proposition on its head: streaming is even more convenient than hopping in your car, but only a subset of content (ever-expanding, to be sure) is on Netflix. That has been more than enough to fuel Netflix’s growth; the service had 222 million subscribers at the end of 2021.

Still, as The Information noted, that number isn’t increasing as quickly as it used to. Netflix sported over 20% year-over-year subscriber growth for years (usually more than that), but hasn’t broken the 20% mark since Q4 2020; growth for the last three quarters was in the single digits. Some of that is likely due to growth that was pulled forward by the pandemic:

Netflix subscriber additions by year

The bigger problem, though, is saturation: Netflix has 75 million subscribers in the US and Canada, where there are around 132 million households. That is nearly as many subscribers as linear TV (84 million), and once you consider shared passwords, penetration may be higher. Other markets like India have more room to grow, but much lower household incomes, and Netflix’s relatively high prices have been an obstacle.

Netflix has ways to grow other than subscribers, most obviously by raising prices. The company has done just that on a mostly annual basis for eight years: in the U.S. the price of a Standard subscription (HD, 2 screens) has increased from $7.99 to $15.49. Netflix executives argue that customers don’t mind because Netflix keeps increasing the amount of content they find compelling; it’s an argument that is easier to accept when subscriber growth is up-and-to-the-right. Now the task is to keep raising prices while ensuring subscriber numbers don’t start going in the opposite direction.

Netflix’s New Initiative: Gaming

To accomplish this Netflix is not only continuing to invest in original programming, but also branching out into new kinds of content, including games. This may seem an odd idea at first: sure, Netflix is generating some new IP, but it would generally be much easier to license that IP than to become proficient at gaming. Netflix, though, believes it has a unique advantage when it comes to gaming: its business model. Chief Product Officer Greg Peters said in the company’s Q2 2021 earnings interview:

Our subscription model yields some opportunities to focus on a set of game experiences that are currently underserved by the sort of dominant monetization models and games. We don’t have to think about ads. We don’t have to think about in-game purchases or other monetization. We don’t have to think about per-title purchases. Really, we can do what we’ve been doing on the movie and series side, which is just hyper laser-focused on delivering the most entertaining game experiences that we can. So we’re finding that many game developers really like that concept and that focus and this idea of being able to put all of their creative energy into just great gameplay and not having to worry about those other considerations that they have typically had to trade off with just making compelling games.

Netflix’s gaming efforts to date have been fairly limited; the company launched with five titles in November, but the fact the company has bought three gaming studios suggests a strong appetite for more — at least amongst Netflix executives.

But what about consumers?

Netflix’s Job: TV

Consumers don’t care so much about business models; they have jobs that they want to get done, and the traditional cable bundle used to do a whole bunch of jobs: information gathering, education, sports, story-telling, escapism, background noise, and more. As I noted in The Great Unbundling, these jobs are increasingly done by completely different services: we get news on the Internet, education from YouTube, story-telling from streaming services, etc.

Netflix is obviously one of those streaming services, but the company is also investing in movies (escapism), and is increasingly the default choice when it comes to the under-appreciated “background noise” category: the service has oceans of low-brow content ready to be streamed while you are barely paying attention. This is a big reason why for many people their choice of streaming services is a matter of which service do they subscribe to in addition to Netflix.

Still, all of these jobs are about passively consuming content; from a consumer perspective gaming is something different, in that you are an active participant. To that end, it’s not clear to me why consumers would even think to consider Netflix when it comes to gaming: that’s not what the service’s job is, nor was it the job of the linear TV bundle that Netflix is helping replace.

Netflix’s Market: Attention

Then again, as founder and co-CEO Reed Hastings likes to say, Netflix’s competition is much broader than TV; Hastings wrote in the company’s Q4 letter to shareholders:

In the US, we earn around 10% of television screen time and less than that of mobile screen time. In 2 other countries, we earn a lower percentage of screen time due to lower penetration of our service. We earn consumer screen time, both mobile and television, away from a very broad set of competitors. We compete with (and lose to) Fortnite more than HBO. When YouTube went down globally for a few minutes in October, our viewing and signups spiked for that time. Hulu is small compared to YouTube for viewing time, and they are successful in the US, but non-existent in Canada, which creates a comparison point: our penetration in the two countries is pretty similar. There are thousands of competitors in this highly-fragmented market vying to entertain consumers and low barriers to entry for those with great experiences. Our growth is based on how good our experience is, compared to all the other screen time experiences from which consumers choose. Our focus is not on Disney+, Amazon or others, but on how we can improve our experience for our members.

Hastings’ point was that analysts should not be overly focused on the threat posed by other streaming services; Netflix has been fighting for attention for years. This is correct, by the way: thanks to the Internet everything from television to social networking to gaming can be delivered at zero marginal cost; the only scarce resource is time, which means attention is the only thing that needs to be competed for.

Well, that and money: companies competing for customer money need a way to communicate to customers what they have to sell and why it is compelling; that means advertising, and advertising requires attention. It follows, then, that the most effective business model in the attention economy is advertising: if customers rely on Google or Facebook to navigate the abundance of content that is the result of zero marginal costs, then it is Google and Facebook that are the best-placed to sell effective ads.

Notice, though, the trouble this Internet reality presents to Netflix: if content is abundant and attention is scarce, it’s easier to sell attention than content; Netflix’s business model, though, is the exact opposite.

Netflix’s Differentiation: Unique Content

Netflix, of course, sees this as a differentiator, and for a long time it was: linear TV had commercials, while Netflix had none. Linear TV made you wait for your favorite show, while Netflix gave you entire seasons at once. This was particularly compelling when Netflix had similar content to linear TV: why would you put up with commercials and TV schedules when you could just stream what you wanted to?

However, as more and more content has moved away from TV and to competing streaming services, differentiation is no longer based on the user experience, but rather uniqueness; on-demand no-commercials is no longer unique, but Stranger Things can only be found on Netflix.

Here Netflix’s biggest advantage is the sheer size of its subscriber base: Netflix can, on an absolute basis, pay more than its streaming competitors for the content it wants, even as its per-subscriber cost basis is lower. This advantage is only accentuated the larger Netflix’s subscriber base gets, and the more revenue it makes per subscriber; the user experience of getting to that unique content doesn’t really matter.

All of these factors make a compelling case for Netflix to start building an advertising business.

First, an advertising-supported or subsidized tier would expand Netflix’s subscriber base, which is not only good for the company’s long-term growth prospects, but also competitive position when it comes to acquiring content. This also applies to the company’s recent attempts to crack down on password sharing, and struggles in the developing world: an advertising-based tier is a much more accessible alternative.

Second, advertising would make it easier for Netflix to continue to raise prices: on one hand, it would provide an alternative for marginal customers who might otherwise churn, and on the other hand, it would create a new benefit for those willing to pay (i.e. no advertising for the highest tiers).

Third, advertising is a natural fit for the jobs Netflix does. Sure, customers enjoy watching shows without ads — and again, they can continue to pay for that — but filler TV, which Netflix also specializes in, is just as easily filled with ads.

Above all, though, is the fact that advertising is a great opportunity that aligns with Netflix’s business: while the company once won with a differentiated user experience worth paying for, today Netflix demands scarce attention because of its investment in unique content. That attention can be sold, and should be, particularly as it increases Netflix’s ability to invest in more unique content, and/or charge higher prices to its user base.

This, I will note, is an about face for me; I’ve long been skeptical that Netflix would ever sell advertising, or that they should. The former may still be warranted, particularly in light of Netflix’s gaming initiative. This feels like solipsism: Netflix’s executives think a lot about their business model, so they are looking for growth opportunities that seem to leverage said business model; I’m not convinced, though, that customers appreciate or care about the differentiation that Netflix claims to be leveraging in gaming, whereas they would appreciate lower prices for streaming, and already have the expectation for ads on TV.

Meanwhile, subscriber growth has stalled, even as the advertising market has proven to be much larger than even Google or Facebook can cover. Moreover, the post-ATT world is freeing up more money for the sort of top-of-funnel advertising that would probably be the norm on a Netflix advertising service. In short, the opportunity is there, the product is right, and the business need is pressing in a way it wasn’t previously.

Of course this would be a lot of work, and a big shift in Netflix’s well-defined value proposition; Netflix, though, has made big shifts before: the entire reason why advertising is a possibility is because Netflix is a streamer, not a DVD mailer. In that view a new (additional) business model is just another rung on Netflix’s ladder.

I wrote a follow-up to this Article in this Daily Update.

An Interview with Nvidia CEO Jensen Huang about Manufacturing Intelligence

Posted on

It took a few moments to realize what was striking about the opening video for Nvidia’s GTC conference: the complete absence of humans.

That the video ended with Jensen Huang, the founder and CEO of Nvidia, is the exception that accentuates the takeaway. On the one hand, the theme of Huang’s keynote was the idea of AI creating AI via machine learning; he called the idea “intelligence manufacting”:

None of these capabilities were remotely possible a decade ago. Accelerated computing, at data center scale, and combined with machine learning, has sped up computing by a million-x. Accelerated computing has enabled revolutionary AI models like the transformer, and made self-supervised learning possible. AI has fundamentally changed what software can make, and how you make software. Companies are processing and refining their data, making AI software, becoming intelligence manufacturers. Their data centers are becoming AI factories. The first wave of AI learned perception and inference, like recognizing images, understanding speech, recommending a video, or an item to buy. The next wave of AI is robotics: AI planning actions. Digital robots, avatars, and physical robots will perceive, plan, and act, and just as AI frameworks like TensorFlow and PyTorch have become integral to AI software, Omniverse will be essential to making robotics software. Omniverse will enable the next wave of AI.

We will talk about the next million-x, and other dynamics shaping our industry, this GTC. Over the past decade, Nvidia-accelerated computing delivered a million-x speed-up in AI, and started the modern AI revolution. Now AI will revolutionize all industries. The CUDA libraries, the Nvidia SDKs, are at the heart of accelerated computing. With each new SDK, new science, new applications, and new industries can tap into the power of Nvidia computing. These SDKs tackle the immense complexity at the intersection of computing, algorithms, and science. The compound effect of Nvidia’s full-stack approach resulted in a million-x speed-up. Today, Nvidia accelerates millions of developers, and tens of thousands of companies and startups. GTC is for all of you.

The core idea behind machine learning is that computers, presented with massive amounts of data, can extract insights and ideas from that data that no human ever could; to put it another way, the development of not just insights but, going forward, software itself, is an emergent process. Nvidia’s role is making massively parallel computing platforms that do the calculations necessary for this emergent process far more quickly than was ever possible with general purpose computing platforms like those undergirding the PC or smartphone.

What is so striking about Nvidia generally and Huang in particular, though, is the extent to which this capability is the result of the precise opposite of an emergent process: Nvidia the company feels like a deliberate design, nearly 29 years in the making. The company started accelerating defined graphical functions, then invented the shader, which made it possible to program the hardware doing that acceleration. This new approach to processing, though, required new tools, so Nvidia invented them, and has been building on their fully integrated stack ever since.

The deliberateness of Nvidia’s vision is one of the core themes I explored in this interview with Huang recorded shortly after his GTC keynote. We also touch on Huang’s background, including immigrating to the United States as a child, Nvidia’s failed ARM acquisition, and more. One particularly striking takeaway for me came at the end of the interview, where Huang said:

Intelligence is the ability to recognize patterns, recognize relationships, reason about it and make a prediction or plan an action. That’s what intelligence is. It has nothing to do with general intelligence, intelligence is just solving problems. We now have the ability to write software, we now have the ability to partner with computers to write software, that can solve many types of intelligence, make many types of predictions at scales and at levels that no humans can.

For example, we know that there are a trillion things on the Internet and the number things on the Internet is large and expanding incredibly fast, and yet we have this little tiny personal computer called a phone, how do we possibly figure out of the trillion things in the internet what we want to see on our little tiny phone? Well, there needs to be a filter in between, what people call the personalized internet, but basically an AI, a recommender system. A recommender that figures out based on the nature of the content, the characteristics of the content, the features of the content, based on your implicit and your explicit and implicit preferences, find a way through all of that to predict what you would like to see. I mean, that’s a miracle! That’s really quite a miracle to be able to do that at scale for everything from movies and books and music and news and videos and you name it, products and things like that. To be able to predict what Ben would want to see, predict what you would want to click on, predict what is useful to you. I’m talking about things that are consumer oriented stuff, but in the future it’ll be predict what is the best financial strategy for you, predict what is the best medical therapy for you, predict what is the best health regimen for you, what’s the best vacation plan for you. All of these things are going to be possible with AI.

As I note in the interview, this should ring a bell for Stratechery readers: what Huang is describing is the computing functionality that undergirds Aggregation Theory, wherein value in a world of abundance accrues to those entities geared towards discovery and providing means of navigating this world that is fundamentally disconnected from the constraints of physical goods and geography. Nvidia’s role in this world is to provide the hardware capability for Aggregation, to be the Intel to Aggregators’ Windows. That, needless to say, is an attractive position to be; like many such attractive positions, it is one that was built not in months or years, but decades.

Read the full interview with Huang here.