As a reminder, each week, every Friday, we’re sending out this overview of content in the Stratechery bundle; highlighted links are free for everyone. Additionally, you have complete control over what we send to you. If you don’t want to receive This Week in Stratechery emails (there is no podcast), please uncheck the box in your delivery settings.
On that note, here were a few of our favorites this week.
Rumors of Google’s Demise… Since the arrival of ChatGPT two-and-a-half years ago, no one in big tech has faced more skepticism than Google. And, looking back, you can make the case that the company actually underinvested in AI infrastructure, which fits the thesis. That, however, is no longer the case: Google (and its new CFO) increased its capital expenditure plans for the second time in six months and Ben is getting pretty bullish. —Andrew Sharp
Content and Community. It is useful to study how the Internet impacted the content industry, because that was a harbinger for how the Internet would affect all industries. AI, though, reduces content to mere tokens, a total commodity. Is there a market there? In Content and Community I argue that any publisher that wants to avoid that fate has to re-anchor itself in the real world: whereas communities and countries created the conditions for successful publishers, publishers of the future needs to create communities with content as the totem. —Ben Thompson
Computers are Entertainment Machines. An enjoyably languid midsummer episode of Dithering begins with discussion of old man ergonomic concerns and Ben’s dorky epiphany for the ideal travel workstation, but in addition to the late-July riffing, the subsequent conversation about AI and Grok yields a trenchant observation on the history of computing in the modern era. To paraphrase: Apple has always styled itself as a tool-making company enabling productivity, but it wasn’t until the iPod and the iPhone—personal consumption devices that have been primarily used for entertainment—that Apple was transformed from a plucky player on the margins into the most successful hardware business in history. It’s a note worth keeping in mind as the whole world focuses on AI for productivity and tries to trace the shape of a market for individuals and businesses looking to enhance efficiency. Historically speaking, the biggest market in tech has always been entertainment, and whoever uses AI to crack that market will likely be the biggest winner of all. —AS
Stratechery Articles and Updates
Content and Community— The old model for content sprung from geographic communities; the new model for content is to be the organizing principle for virtual communities.
Netflix Earnings, Apple and F1— Netflix advertising will change the service; then, F1 might be headed to Apple TV, and it might work.
One of the oldest and most fruitful topics on Stratechery has been the evolution of the content industry, for two reasons: first, it undergirded the very existence of Stratechery itself, which I’ve long viewed not simply as a publication but also as a model for a (then) new content business model.
Second, I have long thought that what happened to content was a harbinger for what would happen to industries of all types. Content was trivially digitized, which means the forces of digital — particularly zero marginal cost reproduction and distribution — manifested in content industries first, but were by no means limited to them. That meant that if you could understand how the Internet impacted publishing — newspapers, books, magazines, music, movies, etc. — you might have a template for what would happen to other industries as they themselves digitized.
AI is the apotheosis of this story and, in retrospect, it’s a story the development of which stretches back not just to the creation of the Internet, but hundreds of years prior and the invention of the printing press. Or, if you really want to get crazy, to the evolution of humanity itself.
The AI Unbundling and Content Commoditization
In September 2022, two months before the release of ChatGPT, I wrote about The AI Unbundling, and traced the history of communication to those ancient times:
As much as newspapers may rue the Internet, their own business model — and my paper delivery job — were based on an invention that I believe is the only rival for the Internet’s ultimate impact: the printing press. Those two inventions, though, are only two pieces of the idea propagation value chain. That value chain has five parts:
The evolution of human communication has been about removing whatever bottleneck is in this value chain. Before humans could write, information could only be conveyed orally; that meant that the creation, vocalization, delivery, and consumption of an idea were all one-and-the-same. Writing, though, unbundled consumption, increasing the number of people who could consume an idea.
Now the new bottleneck was duplication: to reach more people whatever was written had to be painstakingly duplicated by hand, which dramatically limited what ideas were recorded and preserved. The printing press removed this bottleneck, dramatically increasing the number of ideas that could be economically distributed:
The new bottleneck was distribution, which is to say this was the new place to make money; thus the aforementioned profitability of newspapers. That bottleneck, though, was removed by the Internet, which made distribution free and available to anyone.
What remains is one final bundle: the creation and substantiation of an idea. To use myself as an example, I have plenty of ideas, and thanks to the Internet, the ability to distribute them around the globe; however, I still need to write them down, just as an artist needs to create an image, or a musician needs to write a song. What is becoming increasingly clear, though, is that this too is a bottleneck that is on the verge of being removed.
It’s a testament to how rapidly AI has evolved that this observation already feels trite; while I have no idea how to verify these numbers, it seems likely that AI has substantiated more content in the last three years than was substantiated by all of humanity in all of history previously. We have, in other words, reached total content commoditization: the chatbot of your choice will substantiate any content you want on command.
Copyright and Transformation
Many publishers are, as you might expect, up in arms about this reality, and have pinned their hopes for survival on the courts and copyright law. After all, the foundation for all of that new content is the content that came before — content that was created by humans.
The fundamental problem for publishers, however, is that all of this new content is, at least in terms of a textual examination of output, new; in other words, AI companies are soundly winning the first factor of the fair use test, which is whether or not their output is transformative. Judge William Alsup wrote in a lawsuit against Anthropic:
The purpose and character of using copyrighted works to train LLMs to generate new text was quintessentially transformative. Like any reader aspiring to be a writer, Anthropic’s LLMs trained upon works not to race ahead and replicate or supplant them — but to turn a hard corner and create something different. If this training process reasonably required making copies within the LLM or otherwise, those copies were engaged in a transformative use. The first factor favors fair use for the training copies.
There is no serious question that Meta’s use of the plaintiffs’ books had a “further purpose” and “different character” than the books — that it was highly transformative. The purpose of Meta’s copying was to train its LLMs, which are innovative tools that can be used to generate diverse text and perform a wide range of functions. Users can ask Llama to edit an email they have written, translate an excerpt from or into a foreign language, write a skit based on a hypothetical scenario, or do any number of other tasks. The purpose of the plaintiffs’ books, by contrast, is to be read for entertainment or education.
The two judges differed in their view of the fourth factor — the impact that LLMs would have on the market for the copyright holders — but ultimately came to the same conclusion: Judge Alsup said that the purpose of copyright law wasn’t to protect authors from competition for new content, while Judge Chabria said that the authors hadn’t produced evidence of harm.
In fact, I think that both are making the same point (see my earlier analysis here and here): Judge Chabria clearly wished that he could rule in favor of the authors, but to do so would require proving a negative — sales that didn’t happen because would-be customers used LLMs instead. That’s something that seems impossible to ascertain, which gives credence to Judge Alsup’s more simplistic analogy of an LLM to a human author who learned from the books they read. Yes, AI is of such a different scale as to be another category entirely, but given the un-traceability of sales that didn’t happen, the analogy holds for legal purposes.
Publishing’s Three Eras
Still, just because it is impossible to trace specific harm, doesn’t mean harm doesn’t exist. Look no further than the aforementioned history of publishing. To briefly compress hundreds of years of history into three periods:
In the Middle Ages the principal organizing entity for Europe was the Catholic Church. Relatedly, the Catholic Church also held a de facto monopoly on the distribution of information: most books were in Latin, copied laboriously by hand by monks. There was some degree of ethnic affinity between various members of the nobility and the commoners on their lands, but underneath the umbrella of the Catholic Church were primarily independent city-states.
The printing press changed all of this. Suddenly Martin Luther, whose critique of the Catholic Church was strikingly similar to Jan Hus 100 years earlier, was not limited to spreading his beliefs to his local area (Prague in the case of Hus), but could rather see those beliefs spread throughout Europe; the nobility seized the opportunity to interpret the Bible in a way that suited their local interests, gradually shaking off the control of the Catholic Church.
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 language 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 printing press created culture, which itself became the common substrate for nation-states.
Copyright and Franchises
It was nation-states, meanwhile, that made publishing into an incredible money-maker. The most important event in common-law countries was The Statute of Anne in 1710. For the first time the Parliament of Great Britain established the concept of copyright, vested in authors for a limited period of time (14 years, with the possibility of a 14 year renewal); the goal, clearly stated in the preamble, was to incentivize creation:
Whereas printers, booksellers, and other persons have of late frequently taken the liberty of printing, reprinting, and publishing, or causing to be printed, reprinted, and published, books and other writings, without the consent of the authors or proprietors of such books and writings, to their very great detriment, and too often to the ruin of them and their families: for preventing therefore such practices for the future, and for the encouragement of learned men to compose and write useful books; may it please your Majesty, that it may be enacted, and be it enacted by the Queen’s most excellent majesty, by and with the advice and consent of the lords spiritual and temporal, and commons, in this present parliament assembled, and by the authority of the same…
A quarter of a century later America’s founding fathers would, for similar motivations, and in line with the English tradition that undergirded the United States, put copyright in the Constitution:
[The Congress shall have power] To promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries.
These are noble goals; at the same time, it’s important to keep in mind that copyright is an economic distortion, because it is a government-granted monopoly. That, by extension, meant that there was a lot of money to be made in publishing if you could leverage these monopoly rights to your advantage. To focus on the U.S.:
The mid-1800s, led by Benjamin Day and James Gordon Bennett Sr., saw the rise of advertising as a funding source of newspapers, which dramatically decreased the price of an individual copy, expanding reach, which attracted more advertisers.
The turn of the century brought nationwide scale to bear, as entrepreneurs like Joseph Pulitzer and William Randolph Hearst built out nation-wide publishing empires with scaled advertising and reporting.
In the mid-20th century Henry Luce and Condé Montrose Nast created and perfected the magazine model, which combined scale on the back-end with segmentation and targeting on the front-end.
The success of these publishing empires was, in contrast to the first era of publishing, downstream from the existence of nation-states: the fact that the U.S. was a massive market created the conditions for publishing’s golden era and companies that were franchises. That’s Warren Buffett’s term, from a 1991 letter to shareholders:
An economic franchise arises from a product or service that: (1) is needed or desired; (2) is thought by its customers to have no close substitute and; (3) is not subject to price regulation. The existence of all three conditions will be demonstrated by a company’s ability to regularly price its product or service aggressively and thereby to earn high rates of return on capital. Moreover, franchises can tolerate mis-management. Inept managers may diminish a franchise’s profitability, but they cannot inflict mortal damage.
In contrast, “a business” earns exceptional profits only if it is the low-cost operator or if supply of its product or service is tight. Tightness in supply usually does not last long. With superior management, a company may maintain its status as a low-cost operator for a much longer time, but even then unceasingly faces the possibility of competitive attack. And a business, unlike a franchise, can be killed by poor management.
Until recently, media properties possessed the three characteristics of a franchise and consequently could both price aggressively and be managed loosely. Now, however, consumers looking for information and entertainment (their primary interest being the latter) enjoy greatly broadened choices as to where to find them. Unfortunately, demand can’t expand in response to this new supply: 500 million American eyeballs and a 24-hour day are all that’s available. The result is that competition has intensified, markets have fragmented, and the media industry has lost some — though far from all — of its franchise strength.
Given that Buffett wrote this in 1991, he was far more prescient than he probably realized, because the Internet was about to destroy the whole model.
The Internet and Aggregators
The great revelation of the Internet is that copyright wasn’t the only monopoly that mattered to publishers: newspapers in particular benefited from being de facto geographic monopolies as well. The largest newspaper in a particular geographic area attracted the most advertisers, which gave them the most resources to have the best content, further cementing their advantages and the leverage they had on their fixed costs (printing presses, delivery, and reporters). I explained what happened next in 2014’s Economic Power in the Age of Abundance:
One of the great paradoxes for newspapers today is that their financial prospects are inversely correlated to their addressable market. Even as advertising revenues have fallen off a cliff — adjusted for inflation, ad revenues are at the same level as the 1950s — newspapers are able to reach audiences not just in their hometowns but literally all over the world.
The problem for publishers, though, is that the free distribution provided by the Internet is not an exclusive. It’s available to every other newspaper as well. Moreover, it’s also available to publishers of any type, even bloggers like myself.
To be clear, this is absolutely a boon, particularly for readers, but also for any writer looking to have a broad impact. For your typical newspaper, though, the competitive environment is diametrically opposed to what they are used to: instead of there being a scarce amount of published material, there is an overwhelming abundance. More importantly, this shift in the competitive environment has fundamentally changed just who has economic power.
In a world defined by scarcity, those who control the scarce resources have the power to set the price for access to those resources. In the case of newspapers, the scarce resource was reader’s attention, and the purchasers were advertiser…The Internet, though, is a world of abundance, and there is a new power that matters: the ability to make sense of that abundance, to index it, to find needles in the proverbial haystack. And that power is held by Google.
Google was an Aggregator, and publishers — at least those who users visited via a search results page — were a commodity; it was inevitable that money from advertisers in particular would increasingly flow to the former at the expense of the latter.
There were copyright cases against Google, most notably 2006’s Field v. Google, which held that Google’s usage of snippets of the plaintiff’s content was fair use, and furthermore, that Blake Fields, the author, had implicitly given Google a license to cache his content by not specifying that Google not crawl his website.
The crucial point to make about this case, however, and Google’s role on the Internet generally, is that Google posting a snippet of content was good for publishers, at least compared to the AI alternative.
Cloudflare and the AI Content Market
Go back to the two copyright rulings I referenced above: both judges emphasized that the LLMs in question (Claude and Llama) were not reproducing the copyrighted content they were accused of infringing; rather, they were generating novel new content by predicting tokens. Here’s Judge Alsup on how Anthropic used copyrighted work:
Each cleaned copy was translated into a “tokenized” copy. Some words were “stemmed” or “lemmatized” into simpler forms (e.g., “studying” to “study”). And, all characters were grouped into short sequences and translated into corresponding number sequences or “tokens” according to an Anthropic-made dictionary. The resulting tokenized copies were then copied repeatedly during training. By one account, this process involved the iterative, trial-and-error discovery of contingent statistical relationships between each word fragment and all other word fragments both within any work and across trillions of word fragments from other copied books, copied websites, and the like.
Judge Chabria explained how these tokens contribute to the final output:
LLMs learn to understand language by analyzing relationships among words and punctuation marks in their training data. The units of text — words and punctuation marks — on which LLMs are trained are often referred to as “tokens.” LLMs are trained on an immense amount of text and thereby learn an immense amount about the statistical relationships among words. Based on what they learned from their training data, LLMs can create new text by predicting what words are most likely to come next in sequences. This allows them to generate text responses to basically any user prompt.
This isn’t just commoditization: it’s deconstruction. To put it another way, publishers were better off when an entity like Google was copying their text; Google summarizing information — which is what happens with LLM-powered AI Search Overviews — is much worse, even if it’s even less of a copyright violation.
This was a point made to me by Cloudflare CEO Matthew Prince in a conversation we had after I wrote last week about the company’s audacious decision to block AI crawlers on Cloudflare-protected sites by default. What the company is proposing to build is a new model of monetization for publishers; Prince wrote in a blog post:
We’ll work on a marketplace where content creators and AI companies, large and small, can come together. Traffic was always a poor proxy for value. We think we can do better. Let me explain. Imagine an AI engine like a block of swiss cheese. New, original content that fills one of the holes in the AI engine’s block of cheese is more valuable than repetitive, low-value content that unfortunately dominates much of the web today. We believe that if we can begin to score and value content not on how much traffic it generates, but on how much it furthers knowledge — measured by how much it fills the current holes in AI engines “swiss cheese” — we not only will help AI engines get better faster, but also potentially facilitate a new golden age of high-value content creation. We don’t know all the answers yet, but we’re working with some of the leading economists and computer scientists to figure them out.
Cloudflare is calling its initial idea pay per crawl:
Pay per crawl, in private beta, is our first experiment in this area. Pay per crawl integrates with existing web infrastructure, leveraging HTTP status codes and established authentication mechanisms to create a framework for paid content access. Each time an AI crawler requests content, they either present payment intent via request headers for successful access (HTTP response code 200), or receive a 402 Payment Required response with pricing. Cloudflare acts as the Merchant of Record for pay per crawl and also provides the underlying technical infrastructure…
At its core, pay per crawl begins a technical shift in how content is controlled online. By providing creators with a robust, programmatic mechanism for valuing and controlling their digital assets, we empower them to continue creating the rich, diverse content that makes the Internet invaluable…The true potential of pay per crawl may emerge in an agentic world. What if an agentic paywall could operate entirely programmatically? Imagine asking your favorite deep research program to help you synthesize the latest cancer research or a legal brief, or just help you find the best restaurant in Soho — and then giving that agent a budget to spend to acquire the best and most relevant content. By anchoring our first solution on HTTP response code 402, we enable a future where intelligent agents can programmatically negotiate access to digital resources.
I think there is value in Cloudflare’s efforts, which are very much inline with what I proposed in May’s The Agentic Web and Original Sin:
What is possible — not probable, but at least possible — is to in the long run build an entirely new marketplace for content that results in a new win-win-win equilibrium.
First, the protocol layer should have a mechanism for payments via digital currency, i.e. stablecoins. Second, AI providers like ChatGPT should build an auction mechanism that pays out content sources based on the frequency with which they are cited in AI answers. The result would be a new universe of creators who will be incentivized to produce high quality content that is more likely to be useful to AI, competing in a marketplace a la the open web; indeed, this would be the new open web, but one that operates at even greater scale than the current web given the fact that human attention is a scarce resource, while the number of potential agents is infinite.
I do think that there is a market to be made in producing content for AI; it seems likely to me, however, that this market will not save existing publishers. Rather, just as Google created an entirely new class of content sites, Amazon and Meta an entirely new class of e-commerce merchants, and Apple and Meta an entirely new class of app builders, AI will create an entirely new class of token creators who explicitly produce content for LLMs. Existing publishers will participate in this market, but won’t be central to it.
Consider Meta’s market-making as an example. From 2020’s Apple and Facebook:
This explains why the news about large CPG companies boycotting Facebook is, from a financial perspective, simply not a big deal. Unilever’s $11.8 million in U.S. ad spend, to take one example, is replaced with the same automated efficiency that Facebook’s timeline ensures you never run out of content. Moreover, while Facebook loses some top-line revenue — in an auction-based system, less demand corresponds to lower prices — the companies that are the most likely to take advantage of those lower prices are those that would not exist without Facebook, like the direct-to-consumer companies trying to steal customers from massive conglomerates like Unilever.
In this way Facebook has a degree of anti-fragility that even Google lacks: so much of its business comes from the long tail of Internet-native companies that are built around Facebook from first principles, that any disruption to traditional advertisers — like the coronavirus crisis or the current boycotts — actually serves to strengthen the Facebook ecosystem at the expense of the TV-centric ecosystem of which these CPG companies are a part.
In short, Meta advertising made Meta advertisers; along those lines, the extent to which Cloudflare or anyone else manages to create a market for AI content is the extent to which I expect new companies to dominate that market; existing publishers will be too encumbered by their existing audiences and business models — decrepit though it may be — to effectively compete with these new entrants.
Content-Based Communities
So, are existing publishers doomed?
Well, by-and-large yes, but that’s because they have been doomed for a long time. People using AI instead of Google — or Google using AI to provide answers above links — make the long-term outlook for advertising-based publishers worse, but that’s an acceleration of a demise that has been in motion for a long time.
To that end, the answer for publishers in the age of AI is no different than it was in the age of Aggregators: build a direct connection with readers. This, by extension, means business models that maximize revenue per user, which is to say subscriptions (the business model that undergirds this site, and an increasing number of others).
What I think is intriguing, however, is the possibility to go back to the future. Once upon a time publishing made countries; the new opportunity for publishing is to make communities. This is something that AI, particularly as it manifests today, is fundamentally unsuited to: all of that content generated by LLMs is individualized; what you ask, and what the AI answers, is distinct from what I ask, and what answers I receive. This is great for getting things done, but it’s useless for creating common ground.
Stratechery, on the other hand, along with a host of other successful publications, has the potential to be a totem pole around which communities can form. Here is how Wikipedia defines a totem pole:
The word totem derives from the Algonquian word odoodem [oˈtuːtɛm] meaning “(his) kinship group”. The carvings may symbolize or commemorate ancestors, cultural beliefs that recount familiar legends, clan lineages, or notable events. The poles may also serve as functional architectural features, welcome signs for village visitors, mortuary vessels for the remains of deceased ancestors, or as a means to publicly ridicule someone. They may embody a historical narrative of significance to the people carving and installing the pole. Given the complexity and symbolic meanings of these various carvings, their placement and importance lies in the observer’s knowledge and connection to the meanings of the figures and the culture in which they are embedded. Contrary to common misconception, they are not worshipped or the subject of spiritual practice.
The digital environment, thanks in part to the economics of targeted advertising, the drive for engagement, and most recently, the mechanisms of token prediction, is customized to the individual; as LLMs consume everything, including advertising-based media — which, by definition, is meant to be mass market — the hunger for something shared is going to increase.
We already have a great example of this sort of shared experience in sports. Sports, for most people, is itself a form of content: I don’t play football or baseball or basketball or drive an F1 car, but I relish the fact that people around me watch the same games and races that I do, and that that shared experience gives me a reason to congregate and commune with others, and is an ongoing topic of discussion.
Indeed, this desire for a communal topic of interest is probably a factor in the inescapable reach of politics, particularly what happens in Washington D.C.: of course policies matter, but there is an aspect of politics’ prominence that I suspect is downstream of politics as entertainment, and a sorting mechanism for community.
In short, there is a need for community, and I think content, whether it be an essay, a podcast, or a video, can be artifacts around which communities can form and sustain themselves, ultimately to the economic benefit of the content creator. There is, admittedly, a lot to figure out in terms of that last piece, but when you remember that content made countries, the potential upside is likely quite large indeed.
One of the most paradoxical aspects of AI is that while it is hailed as the route to abundance, the most important financial outcomes have been about scarcity. The first and most obvious example has been Nvidia, whose valuation has skyrocketed while demand for its chips continues to outpace supply:
Another scarce resource that has come to the forefront over the last few months is AI talent; the people who are actually building and scaling the models are suddenly being paid more than professional athletes, and it makes sense:
The potential financial upside from “winning” in AI are enormous
Outputs are somewhat measurable
The work-to-be-done is the same across the various companies bidding for talent
It’s that last point that is fairly unique in tech history. While great programmers have always been in high demand, and there have been periods of intense competition in specific product spaces, over the past few decades tech companies have been franchises, wherein their market niches have been fairly differentiated: Google and search, Amazon and e-commerce, Meta and social media, Microsoft and business applications, Apple and devices, etc. This reality meant that the company mattered more than any one person, putting a cap on individual contributor salaries.
AI, at least to this point, is different: in the long run it seems likely that there will be dominant product companies in various niches, but as long as the game is foundational models, then everyone is in fact playing the same game, which elevates the bargaining power of the best players. It follows, then, that the team they play for is the team that pays the most, through some combination of money and mission; by extension, the teams that are destined to lose are the ones who can’t or won’t offer enough of either.
Apple’s Reluctance
It’s that last point I’m interested in; I’m not in position to judge the value of any of the players changing teams, but the teams are worth examining. Consider Meta and Apple and the latest free agent signing; from Bloomberg:
Apple Inc.’s top executive in charge of artificial intelligence models is leaving for Meta Platforms Inc., another setback in the iPhone maker’s struggling AI efforts. Ruoming Pang, a distinguished engineer and manager in charge of the company’s Apple foundation models team, is departing, according to people with knowledge of the matter. Pang, who joined Apple from Alphabet Inc. in 2021, is the latest big hire for Meta’s new superintelligence group, said the people, who declined to be named discussing unannounced personnel moves.
To secure Pang, Meta offered a package worth tens of millions of dollars per year, the people said. Meta Chief Executive Officer Mark Zuckerberg has been on a hiring spree, bringing on major AI leaders including Scale AI’s Alexandr Wang, startup founder Daniel Gross and former GitHub CEO Nat Friedman with high compensation. Meta has also hired Yuanzhi Li, a researcher from OpenAI, and Anton Bakhtin, who worked on Claude at Anthropic PBC, according to other people with knowledge of the matter. Last month, it hired a slew of other OpenAI researchers. Meta, later on Monday, confirmed it is hiring Pang. Apple, Pang, OpenAI and Anthropic didn’t respond to requests for comment.
That Apple is losing AI researchers is a surprise only in that they had researchers worth hiring; after all, this is the company who already implicitly signaled its AI reluctance in terms of that other scarce resource: Nvidia chips. Again from Bloomberg:
Former Chief Financial Officer Luca Maestri’s conservative stance on buying GPUs, the specialized circuits essential to AI, hasn’t aged well either. Under Cook, Apple has used its market dominance and cash hoard to shape global supply chains for everything from semiconductors to the glass for smartphone screens. But demand for GPUs ended up overwhelming supply, and the company’s decision to buy them slowly — which was in line with its usual practice for emerging technologies it isn’t fully sold on — ended up backfiring. Apple watched as rivals such as Amazon and Microsoft Corp. bought much of the world’s supply. Fewer GPUs meant Apple’s AI models were trained all the more slowly. “You can’t magically summon up more GPUs when the competitors have already snapped them all up,” says someone on the AI team.
It may seem puzzling that the company that in its 2024 fiscal year generated $118 billion in free cash flow would be so cheap, but Apple’s reluctance makes sense from two perspectives.
First, the potential impact of AI on Apple’s business prospects, at least in the short term, are fairly small: we still need devices on which to access AI, and Apple continues to own the high end of devices (there is, of course, long-term concern about AI obviating the need for a smartphone, or meaningfully differentiating an alternative platform like Android). That significantly reduces the financial motivation for Apple to outspend other companies in terms of both GPUs and researchers.
Second, AI, at least some of the more fantastical visions painted by companies like Anthropic, is arguably counter to Apple’s entire ethos as a company.
Tech’s Two Philosophies
It was AI, at least the pre-LLM version of it, that inspired me in 2018 to write about Tech’s Two Philosophies; one was represented by Google and Facebook (now Meta):
In Google’s view, computers help you get things done — and save you time — by doing things for you. Duplex was the most impressive example — a computer talking on the phone for you — but the general concept applied to many of Google’s other demonstrations, particularly those predicated on AI: Google Photos will not only sort and tag your photos, but now propose specific edits; Google News will find your news for you, and Maps will find you new restaurants and shops in your neighborhood. And, appropriately enough, the keynote closed with a presentation from Waymo, which will drive you…
Zuckerberg, as so often seems to be the case with Facebook, comes across as a somewhat more fervent and definitely more creepy version of Google: not only does Facebook want to do things for you, it wants to do things its chief executive explicitly says would not be done otherwise. The Messianic fervor that seems to have overtaken Zuckerberg in the last year, though, simply means that Facebook has adopted a more extreme version of the same philosophy that guides Google: computers doing things for people.
The other philosophy was represented by Apple and Microsoft:
Earlier this week, while delivering Microsoft’s Build conference keynote, CEO Satya Nadella struck a very different tone…This is technology’s second philosophy, and it is orthogonal to the other: the expectation is not that the computer does your work for you, but rather that the computer enables you to do your work better and more efficiently. And, with this philosophy, comes a different take on responsibility. Pichai, in the opening of Google’s keynote, acknowledged that “we feel a deep sense of responsibility to get this right”, but inherent in that statement is the centrality of Google generally and the direct culpability of its managers. Nadella, on the other hand, insists that responsibility lies with the tech industry collectively, and all of us who seek to leverage it individually.
This second philosophy, that computers are an aid to humans, not their replacement, is the older of the two; its greatest proponent — prophet, if you will — was Microsoft’s greatest rival, and his analogy of choice was, coincidentally enough, about transportation as well. Not a car, but a bicycle:
I remember reading an article when I was about 12 years old, I think it might have been in Scientific American, where they measured the efficiency of locomotion for all these species on planet earth, how many kilocalories did they expand to get from point A to point B, and the condor came in the top of the list, surpassed everything else, and humans came in about a third of the way down the list, which was not such a great showing for the crown of creation.
But somebody there had the imagination to test the efficiency of a human riding a bicycle, and a human riding a bicycle blew away the condor all the way off the top of the list. And it made a really big impression on me that we humans are tool builders, and that we can fashion tools that amplify these inherent abilities that we have to spectacular magnitudes. And so for me a computer has always been a bicycle of the mind, something that takes us far beyond our inherent abilities. I think we’re just at the early stages of this tool, very early stages, and we’ve come only a very short distance, and it’s still in its formation, but already we’ve seen enormous changes. I think that’s nothing compared to what’s coming in the next 100 years.
We are approximately forty years on from that clip, and Steve Jobs’ prediction that enormous changes were still to come is obviously prescient: mobile and the Internet have completely transformed the world, and AI is poised to make those impacts look like peanuts. What I’m interested in in the context of this Article, however, is the interplay between business opportunity — or risk — and philosophy. Apple’s position is here:
In this view the company’s conservatism makes sense: Apple doesn’t quite see the upside of AI for their business (and isn’t overly concerned about the downsides), and its bias towards tools means that AI apps on iPhones are sufficient; Apple might be an increasingly frustrating platform steward, but they are at their core a platform company, and apps on their platform are delivering Apple users AI tools.
This same framework also explains Meta’s aggressiveness. First, the opportunity is huge, as I documented last fall in Meta’s AI Abundance (and, for good measure, there is risk as well, as time — the ultimate scarcity for an advertising-based business — is spent using AI). Second, Meta’s philosophy is that computers do things for you:
Given this graph, is it any surprise that Meta hired away Apple’s top AI talent?
I’m Feeling Lucky
Another way to think about how companies are approaching AI is through the late Professor Clayton Christensen’s discussion around sustaining versus disruptive innovation. From an Update last month after the news of Meta’s hiring spree first started making waves:
The other reason to believe in Meta versus Google comes down to the difference between disruptive and sustaining innovations. The late Professor Clayton Christensen described the difference in The Innovator’s Dilemma:
Most new technologies foster improved product performance. I call these sustaining technologies. Some sustaining technologies can be discontinuous or radical in character, while others are of an incremental nature. What all sustaining technologies have in common is that they improve the performance of established products, along the dimensions of performance that mainstream customers in major markets have historically valued. Most technological advances in a given industry are sustaining in character. An important finding revealed in this book is that rarely have even the most radically difficult sustaining technologies precipitated the failure of leading firms.
Occasionally, however, disruptive technologies emerge: innovations that result in worse product performance, at least in the near-term. Ironically, in each of the instances studied in this book, it was disruptive technology that precipitated the leading firms’ failure. Disruptive technologies bring to a market a very different value proposition than had been available previously. Generally, disruptive technologies underperform established products in mainstream markets. But they have other features that a few fringe (and generally new) customers value. Products based on disruptive technologies are typically cheaper, simpler, smaller, and, frequently, more convenient to use.
The question of whether generative AI is a sustaining or disruptive innovation for Google remains uncertain two years after I raised it. Obviously Google has tremendous AI capabilities both in terms of infrastructure and research, and generative AI is a sustaining innovation for its display advertising business and its cloud business; at the same time, the long-term questions around search monetization remain as pertinent as ever.
Meta, however, does not have a search business to potentially disrupt, and a whole host of ways to leverage generative AI across its business; for Zuckerberg and company I think that AI is absolutely a sustaining technology, which is why it ultimately makes sense to spend whatever is necessary to get the company moving in the right direction.
The problem with this analysis is the Google part: how do you square the idea that AI is disruptive to Google with the fact that they are investing just has heavily as everyone else, and in fact started far earlier than everyone else? I think the answer goes back to Google’s founding, and the “I’m Feeling Lucky” button:
While that button is now gone from Google.com, I don’t think it was an accident that it persisted long after it was even usable (instant search results meant that by 2010 you didn’t even have a chance to click it); “I’m Feeling Lucky” was a statement of purpose. From 2016’s Google and the Limits of Strategy:
In yesterday’s keynote, Google CEO Sundar Pichai, after a recounting of tech history that emphasized the PC-Web-Mobile epochs I described in late 2014, declared that we are moving from a mobile-first world to an AI-first one; that was the context for the introduction of the Google Assistant.
It was a year prior to the aforementioned iOS 6 that Apple first introduced the idea of an assistant in the guise of Siri; for the first time you could (theoretically) compute by voice. It didn’t work very well at first (arguably it still doesn’t), but the implications for computing generally and Google specifically were profound: voice interaction both expanded where computing could be done, from situations in which you could devote your eyes and hands to your device to effectively everywhere, even as it constrained what you could do. An assistant has to be far more proactive than, for example, a search results page; it’s not enough to present possible answers: rather, an assistant needs to give the right answer.
This is a welcome shift for Google the technology; from the beginning the search engine has included an “I’m Feeling Lucky” button, so confident was Google founder Larry Page that the search engine could deliver you the exact result you wanted, and while yesterday’s Google Assistant demos were canned, the results, particularly when it came to contextual awareness, were far more impressive than the other assistants on the market. More broadly, few dispute that Google is a clear leader when it comes to the artificial intelligence and machine learning that underlie their assistant.
The problem — apparent even then — was the conflict with Google’s business model:
A business, though, is about more than technology, and Google has two significant shortcomings when it comes to assistants in particular. First, as I explained after this year’s Google I/O, the company has a go-to-market gap: assistants are only useful if they are available, which in the case of hundreds of millions of iOS users means downloading and using a separate app (or building the sort of experience that, like Facebook, users will willingly spend extensive amounts of time in).
Secondly, though, Google has a business-model problem: the “I’m Feeling Lucky” button guaranteed that the search in question would not make Google any money. After all, if a user doesn’t have to choose from search results, said user also doesn’t have the opportunity to click an ad, thus choosing the winner of the competition Google created between its advertisers for user attention. Google Assistant has the exact same problem: where do the ads go?
What I articulated in that Article was Google’s position on this graph:
AI is the ultimate manifestation of “I’m Feeling Lucky”; Google has been pursuing AI because that is why Page and Brin started the company in the first place; business models matter, but they aren’t dispositive, and while that may mean short-term difficulties for Google, it is a reason to be optimistic that the company will figure out AI anyways.
Microsoft, OpenAI, and Anthropic
Frameworks like this are useful, but not fully explanatory; I think this particular one goes a long way towards contextualizing the actions of Apple, Meta, and Google, but is much more speculative for some other relevant AI players. Consider Microsoft, which I would place here:
Microsoft doesn’t have any foundational models of note, but has invested heavily in OpenAI; its most important AI product are its various Copilots, which are indeed a bet on the “tool” philosophy. The question, as I laid out last year in Enterprise Philosophy and the First Wave of AI, is whether rank-and-file employees want Microsoft’s tools:1
Notice, though, how this aligned with the Apple and Microsoft philosophy of building tools: tools are meant to be used, but they take volition to maximize their utility. This, I think, is a challenge when it comes to Copilot usage: even before Copilot came out employees with initiative were figuring out how to use other AI tools to do their work more effectively. The idea of Copilot is that you can have an even better AI tool — thanks to the fact it has integrated the information in the “Microsoft Graph” — and make it widely available to your workforce to make that workforce more productive.
To put it another way, the real challenge for Copilot is that it is a change management problem: it’s one thing to charge $30/month on a per-seat basis to make an amazing new way to work available to all of your employees; it’s another thing entirely — a much more difficult thing — to get all of your employees to change the way they work in order to benefit from your investment, and to make Copilot Pages the “new artifact for the AI age”, in line with the spreadsheet in the personal computer age.
OpenAI’s nascent strength in the enterprise market is giving its partner and biggest investor indigestion. Microsoft salespeople describe being caught flatfooted at a time when they’re under pressure to get Copilot into as many customers’ hands as possible. The behind-the-scenes dogfight is complicating an already fraught relationship between Microsoft and OpenAI…It’s unclear whether OpenAI’s momentum with corporations will continue, but the company recently said it has 3 million paying business users, a 50% jump from just a few months earlier. A Microsoft spokesperson said Copilot is used by 70% of the Fortune 500 and paid users have tripled compared with this time last year…
This story is based on conversations with more than two dozen customers and salespeople, many of them Microsoft employees. Most of these people asked not to be named in order to speak candidly about the competition between Microsoft and OpenAI. Both companies are essentially pitching the same thing: AI assistants that can handle onerous tasks — researching and writing; analyzing data — potentially letting office workers focus on thornier challenges. Since both chatbots are largely based on the same OpenAI models, Microsoft’s salesforce has struggled to differentiate Copilot from the much better-known ChatGPT, according to people familiar with the situation.
As long as AI usage relies on employee volition, ChatGPT has the advantage; what is interesting about this observation, however, is that it shows that OpenAI is actually in the same position as Microsoft:
This, by extension, explains why Anthropic is different; the other leading independent foundational lab is clearly focused on agents, not chatbots, i.e. AI that does stuff for you, instead of a tool. Consider the contrast between Cursor and Claude Code: Cursor is an integrated development environment (IDE) that provides the best possible UI for AI-augmented programming; Claude Code, on the other hand, barely bothers with a UI at all. It runs in the terminal, which people put up with because it is the best at one-shotting outputs; this X thread was illuminating:
More generally, I wrote in an Update after the release of Claude 4, which was heavily focused on agentic workloads:
Computing didn’t start with the personal computer, but rather with the replacement of the back office. Or, to put it in rather more dire terms, the initial value in computing wasn’t created by helping Boomers do their job more efficiently, but rather by replacing entire swathes of them completely…Agents aren’t copilots; they are replacements. They do work in place of humans — think call centers and the like, to start — and they have all of the advantages of software: always available, and scalable up-and-down with demand…
Benioff isn’t talking about making employees more productive, but rather companies; the verb that applies to employees is “augmented”, which sounds much nicer than “replaced”; the ultimate goal is stated as well: business results. That right there is tech’s third philosophy: improving the bottom line for large enterprises.
Notice how well this framing applies to the mainframe wave of computing: accounting and ERP software made companies more productive and drove positive business results; the employees that were “augmented” were managers who got far more accurate reports much more quickly, while the employees who used to do that work were replaced. Critically, the decision about whether or not to make this change did not depend on rank-and-file employees changing how they worked, but for executives to decide to take the plunge.
This strikes me as a very worthwhile goal, at least from a business perspective. OpenAI is busy owning the consumer space, while Google and its best-in-class infrastructure and leading models struggles with product; Anthropic’s task is to build the best agent product in the world, including not just state-of-the-art models but all of the deterministic computing scaffolding that actually makes them replacement-level workers. After all, Anthropic’s API pricing may look expensive relative to Google, but it looks very cheap relative to a human salary.
That means that Anthropic shares the upper-right quadrant with Meta:
Again, this is just one framework; there are others. Moreover, the boundaries are fuzzy. OpenAI is working on agentic workloads, for example, and the hyperscalers all benefit from more AI usage, whether user- or agent-driven; Google, meanwhile, is rapidly evolving Search to incorporate generative AI.
At the same time, to go back to the talent question, I don’t think it’s a surprise that Meta appears to be picking off more researchers from OpenAI than from Anthropic: my suspicion is that to the extent mission is a motivator the more likely an AI researcher is to be enticed by the idea of computers doing everything, instead of merely augmenting humans. And, by extension, the incumbent tool-makers may have no choice but to partner with the true believers.
The story of 2022 was the emergence of AI, first with image generation models, including DALL-E, MidJourney, and the open source Stable Diffusion, and then ChatGPT, the first text-generation model to break through in a major way. It seems clear to me that this is a new epoch in technology.
Sometimes the accuracy of a statement is measured by its banality, and that certainly seems to be the case here: AI is the new epoch, consuming the mindshare of not just Stratechery but also the companies I cover. To that end, two-and-a-half years on, I thought it would be useful to revisit that 2023 analysis and re-evaluate the state of AI’s biggest players, primarily through the lens of the Big Five: Apple, Google, Meta, Microsoft, and Amazon.
The proximate cause for this reevaluation is the apparent five alarm fire that is happening at Meta: the company’s latest Llama 4 release was disappointing — and in at least one case, deceptive — pushing founder and CEO Mark Zuckerberg to go on a major spending spree for talent. From The Wall Street Journal over the weekend:
Mark Zuckerberg is spending his days firing off emails and WhatsApp messages to the sharpest minds in artificial intelligence in a frenzied effort to play catch-up. He has personally reached out to hundreds of researchers, scientists, infrastructure engineers, product stars and entrepreneurs to try to get them to join a new Superintelligence lab he’s putting together…And Meta’s chief executive isn’t just sending them cold emails. Zuckerberg is also offering hundreds of millions of dollars, sums of money that would make them some of the most expensive hires the tech industry has ever seen. In at least one case, he discussed buying a startup outright.
While the financial incentives have been mouthwatering, some potential candidates have been hesitant to join Meta Platforms’ efforts because of the challenges that its AI efforts have faced this year, as well as a series of restructures that have left prospects uncertain about who is in charge of what, people familiar with their views said. Meta’s struggles to develop cutting-edge artificial-intelligence technology reached a head in April, when critics accused the company of gaming a leaderboard to make a recently released AI model look better than it was. They also delayed the unveiling of a new, flagship AI model, raising questions about the company’s ability to continue advancing quickly in an industrywide AI arms race…
For those who have turned him down, Zuckerberg’s stated vision for his new AI superteam was also a concern. He has tasked the team, which will consist of about 50 people, with achieving tremendous advances with AI models, including reaching a point of “superintelligence.” Some found the concept vague or without a specific enough execution plan beyond the hiring blitz, the people said.
That last paragraph complicates analysis generally. In my January 2023 Article I framed my evaluation through Professor Clay Christensen’s framework of sustaining versus disruptive innovation: was AI complementary to existing business models (i.e. Apple devices are better with AI) or disruptive to them (i.e. AI might be better than Search but monetize worse). A higher level question, however, is if AI simply obsoletes everything, from tech business models to all white collar work to work generally or even to life itself.
Perhaps it is the smallness of my imagination or my appreciation of the human condition that makes me more optimistic than many about the probability of the most dire of predictions: I think they are quite low. At the same time, I think that those dismissing AI as nothing but hype are missing the boat as well. This is a big deal, even if the changes may end up fitting into the Bill Gates maxim that “We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.”
To that end, let’s go back two years to AI and the Big Five, and consider where we might be in eight.
Apple
Infrastructure: Minimal Model: None Partner: OpenAI? Data: No public data differentiation, potential private data differentiation Distribution: Apple devices Core Business: Devices are complementary to current AI use cases Scarcity Risk: Could lose differentiation in high-end hardware Disruptive/Sustaining: Sustaining New Business Potential: Robotics
Apple has had a dramatic few years marked by the debacle that has been Apple Intelligence: the company has basic on-device LLM capabilities and its own private cloud compute infrastructure, but is nowhere near the cutting edge in terms of either models or products.
The company’s saving grace, however, is that its core business is not immediately threatened by AI. OpenAI, Claude, etc. are, from a consumer perspective, apps that you use on your iPhone or in a browser; Cursor is an IDE you use on your Mac. Apple’s local LLMs, meanwhile, can potentially differentiate apps built for Apple platforms, and Apple has unique access to consumer data and, by extension, the means to build actually usable and scalable individual semantic indexes over which AI can operate.
This positioning isn’t a panacea; in April’s Apple and the Ghosts of Companies Past, I analogized Apple’s current position to Microsoft and the Internet: everyone used the Internet on Windows PCs, but it was the Internet that created the conditions for the paradigm that would surpass the PC, which was mobile.
What Microsoft did that Intel — another company I compared Apple to — did not, was respond to their mobile miss by accepting their loss, building a complementary business (cloud computing), which then positioned them for the AI paradigm. Apple should do something similar: I am very encouraged by the company’s deepening partnership with OpenAI in iOS 26, and the company should double-down on being the best hardware for what appears to be the dominant consumer AI company.
The way this leads to a Microsoft-like future is by putting the company’s efforts towards building hardware beyond the phone. Yes, OpenAI has acqui-hired Jony Ive and his team of Apple operational geniuses, but Apple should take that as a challenge to provide OpenAI with better hardware and bigger scale than the horizontal services company can build on their own. That should mean a host of AI-powered devices beyond the phone, including the Apple Watch, HomePod, glasses, etc.; in the long run Apple should be heavily investing in robotics and home automation. There is still no one better at consumer hardware than Apple, both in terms of quality and also scalability, and they should double down on that capability.
If Apple does feel the need to go it alone, then the company needs to make a major acquisition and commit to spending billions of dollars. The best option would be Mistral: the company has a lot of talent (including a large portion of the team that built Meta’s early and well-regarded Llama models), and its open source approach is complementary to Apple’s business. It’s unclear, however, if French and European authorities would allow the current jewel of the European startup ecosystem to be acquired by an American company; regardless, Apple needs to either commit to partnering — and the loss of control that entails — or commit to spending a lot more money than they have to date.
Google
Infrastructure: Best Model: Good Partner: None Data: Best Distribution: Android devices, Search, GCP Core Business: Chatbots are disruptive to Search Scarcity Risk: Data feedback loops diminished Disruptive/Sustaining: Disruptive New Business Potential: Cloud
Google is in many respects the opposite of Apple: the search company’s last two years have gone better than I anticipated (while Apple’s have gone worse), but their fundamental position and concerns remain mostly unchanged (in Apple’s case, that’s a good thing; for Google it’s a concern).
Google’s infrastructure is in many respects the best in the world, and by a significant margin. The company is fully integrated from chips to networking to models, and regularly points to that integration as the key to unlocking capabilities like Gemini’s industry-leading context window size; Gemini also has the most attractive pricing of any leading AI model. On the other hand, integration can have downsides: Google’s dependence on its own TPUs means that the company is competing with the Nvidia ecosystem up-and-down the stack; this is good for direct cost savings, but could be incurring hidden costs in terms of tooling and access to innovations.
Gemini, meanwhile, has rapidly improved, and scores very highly in LLM evaluations. There is some question as to whether or not the various model permutations are over-indexed on those LLM valuations; real world usage of Gemini seems to significantly lag behind OpenAI and Anthropic’s respective models. Where Google is undoubtedly ahead is in adjacent areas like media generation; Veo in particular appears to have no peers when it comes to video generation.
This speaks to what might be Google’s most important advantage: data. Veo can draw on YouTube video, the scale of which is hard to fathom. Google’s LLMs, meanwhile, benefit not only from Google’s leading position in terms of indexing the web, but also the fact that no website can afford to block Google’s crawler. Google has also spent years collecting other forms of data, like scanning books, archiving research papers, etc.
Google also has distribution channels, particularly Android: the potential to deliver an integrated device and cloud AI experience is compelling, and is Android’s best chance yet to challenge Apple’s dominance of the high end. Delivering on that integration will be key, however: ChatGPT dominates consumer mindshare to-date, and, as I noted above, Apple can and should differentiate itself as the best devices for using ChatGPT; can Google make its devices better by controlling both the model and the operating system (and access to all of the individual consumer data that entails)? Or, to put it another way, can Google actually make a good product?
The problem for Google, just as one could foresee two years ago (and even years before then), is the disruptive potential of AI for its core business of Search. The problem with having a near perfect business model — which is what Google had with Search ads where users picked the winner of an auction — is that there is nowhere to go but down. Given that, I think that Google has done well with its focus on AI Search Overviews to make Search better and, at least so far, maintain monetization rates; I also think the company’s development of the Search Funnel to systemically evolve search for AI is a smart approach to tilt AI from being disruptive to sustaining.
What is much more promising is cloud computing, which is where Google’s infrastructure and model advantages (particularly in terms of pricing) can truly be brought to bear, without the overhand of needing to preserve revenue or reignite sclerotic product capabilities. Google Cloud Platform has been very focused on fitting into multi-cloud workflows, but the big potential in the long run is that Google’s AI capabilities act as gravity for an ever increasing share of enterprise cloud workflows generally.
Meta
Infrastructure: Good Model: OK Partner: None Data: Good Distribution: Meta apps, Quest devices Core Business: AI delivers more individualized content and ads Scarcity Risk: Attention diverted to chatbots Disruptive/Sustaining: Sustaining New Business Potential: Virtual worlds and generative UIs
Meta’s positioning is somewhere between Apple and Google, but I’ve assumed it was closer to the former; the company may be scuffling more than expected in AI, but its core strategic positioning seems more solid: more individualized content slots into Meta’s distribution channels, and generative ads should enhance Meta’s offering for its long tail advertising base. Generative AI, meanwhile, is very likely the key to Meta realizing a return on its XR investments, both by creating metaverses for VR and UI for AR. I’m on the record as being extremely optimistic about Meta’s AI Abundance.
There are risks, however. The scarce resource Meta competes for is attention, and LLMs are already consuming huge amounts of it and expanding all the time. There is an argument that this makes chatbots just as much of a problem for Meta as they are for Google: even if Meta gets lots of users using Meta AI, time spent using Meta AI is time not spent consuming formats that are better suited to monetization. The difference is that I think that Meta would do a better job of monetizing those new surfaces: there is no expectation of objectivity or reliability to maintain like there is with search; sometimes being a lowest common denominator interface is an asset.
That noted, what does seem clear from Zuckerberg’s spending spree is that these risks are probably bigger than I appreciated. While Zuckerberg has demonstrated with the company’s Reality Labs division that he is willling to invest billions of dollars in an uncertain future, the seeming speed and desperation with these AI recruiting efforts strongly suggests that the company’s core business is threatened in ways I didn’t properly appreciate or foresee two years ago. In retrospect, however, that makes sense: the fact that there are so many upside scenarios for Meta with AI by definition means there is a lot of downside embedded in not getting AI right; the core business of a company like Apple, on the other hand, is sufficiently removed from AI that both its upside and downside are more limited, relatively speaking.
What does concern me is the extent to which Meta seems to lack direction with AI; that was my biggest takeaway from my last interview with Zuckerberg. I felt like I had more ideas for how generative AI could impact the company’s business than Zuckerberg did (Zuckerberg’s comments on the company’s recent earnings call were clearly evolved based on our interview, which took place two days prior); that too fits with the current frenzy. Zuckerberg seems to have belatedly realized not only that the company’s models are falling behind, but that the overall AI effort needs new leadership and new product thinking; thus Alexandr Wang for knowledge on the state of the art, Nat Friedman for team management, and Daniel Gross for product. It’s not totally clear how this team will be organized or function, but what is notable — and impressive, frankly — is the extent to which Zuckerberg is implicitly admitting he has a problem. That’s the sort of humility and bias towards action that Apple could use.
Microsoft
Infrastructure: Very Good Model: None Partner: OpenAI Data: Good Distribution: Windows, Microsoft 365, Azure Core Business: AI drives Azure usage Scarcity Risk: Access to leading edge models Disruptive/Sustaining: Sustaining for Azure, potentially disruptive for Microsoft 365 New Business Potential: Agents
Microsoft’s position seemed unimpeachable in January 2023; this is the entirety of what wrote in AI and the Big Five:
Microsoft, meanwhile, seems the best placed of all. Like AWS it has a cloud service that sells GPUs; it is also the exclusive cloud provider for OpenAI. Yes, that is incredibly expensive, but given that OpenAI appears to have the inside track to being the AI epoch’s addition to this list of top tech companies, that means that Microsoft is investing in the infrastructure of that epoch.
Bing, meanwhile, is like the Mac on the eve of the iPhone: yes it contributes a fair bit of revenue, but a fraction of the dominant player, and a relatively immaterial amount in the context of Microsoft as a whole. If incorporating ChatGPT-like results into Bing risks the business model for the opportunity to gain massive market share, that is a bet well worth making.
The latest report from The Information, meanwhile, is that GPT is eventually coming to Microsoft’s productivity apps. The trick will be to imitate the success of AI-coding tool GitHub Copilot (which is built on GPT), which figured out how to be a help instead of a nuisance (i.e. don’t be Clippy!).
What is important is that adding on new functionality — perhaps for a fee — fits perfectly with Microsoft’s subscription business model. It is notable that the company once thought of as a poster child for victims of disruption will, in the full recounting, not just be born of disruption, but be well-placed to reach greater heights because of it.
Microsoft is still well-positioned, but things are a bit more tenuous than what I wrote in that Article:
Microsoft’s relationship with OpenAI is increasingly frayed, most recently devolving into OpenAI threats of antitrust complaints if Microsoft doesn’t relinquish its rights to future profits and agree to OpenAI’s proposed for-profit restructuring. My read of the situation is that Microsoft still has most of the leverage in the relationship — thus the threats of government involvement — but only until 2030 when the current deal expires.
Bing provided the remarkable Sydney, which Microsoft promptly nerfed; relatedly, Bing appears to have gained little if any ground thanks to its incorporation of AI.
GitHub Copilot has been surpassed by startups like Cursor and dedicated offerings from foundation model makers, while the lack of usage and monetization numbers for Microsoft’s other Copilot products are perhaps telling in their absence.
What is critical — and why I am still bullish on Microsoft’s positioning — is the company’s infrastructure and distribution advantages. I already referenced the company’s pivot after missing mobile: the payoff to The End of Windows was that the company was positioned to capitalize on AI when the opportunity presented itself in a way that Intel was not. Microsoft also, by being relatively late, is the most Nvidia-centric of the hyperscalers; Google is deeply invested in TPUs (although they offer Nvidia instances), while Amazon’s infrastructure is optimized for commodity computing (and is doubling down with Trainium).
Azure, meanwhile, is the exclusive non-OpenAI provider of OpenAI APIs, which not only keeps Microsoft enterprise customers on Azure, but is also a draw in their own right. To that end, I think that Microsoft’s priority in their negotiations with OpenAI should be on securing this advantage for Azure permanently, even if that means giving up many of their rights to OpenAI’s business as a whole.
The other thing that Microsoft should do is deepen their relationship — and scale of investment — in alternative model providers. It was a good sign that xAI CEO Elon Musk appeared in a pre-recorded video at Microsoft Build; Microsoft should follow that up with an investment in helping ensure that xAI continues to pursue the leading edge of AI models. Microsoft has also made small-scale investments in Mistral and should consider helping fund Llama; these sorts of investments are expensive, but not having access to leading models — or risking total dependency on Sam Altman’s whims — would be even pricier.
Amazon
Infrastructure: Good Model: Poor Partner: Anthropic Data: Good Distribution: AWS, Alexa Core Business: AI drives AWS usage Scarcity Risk: Access to leading edge models, chip competitiveness Disruptive/Sustaining: Sustaining for AWS, long tail e-commerce recommendations potentially disruptive for Amazon.com New Business Potential: Agents, Affiliate revenue from AI recommendations on Amazon.com
I’ve become much more optimistic about Amazon’s position over the last two years:
First, there is the fact that AI is not disruptive for any of Amazon’s businesses; if anything, they all benefit. Increased AWS usage is obvious, but also Amazon.com could be a huge beneficiary of customers using AI for product recommendations (on the other hand, AI could be more effective at finding and driving long tail e-commerce alternatives, or reducing the importance of Amazon.com advertising). AWS is also primarily monetized via usage, not seats; to the extent AWS-based seat-driven SaaS companies are disrupted is the extent to which AWS will probably earn more usage revenue from AI-based disruptors.
Second, AWS’s partnership with Anthropic also seems much more stable than Microsoft’s partnership with OpenAI. ChatGPT obviously drives a ton of Azure usage, but it’s also the core reason why OpenAI and Microsoft’s conflict was inevitable; Anthropic’s lack of a strong consumer play means it is much more tenable, if not downright attractive, for them to have a supplier-type of relationship with AWS, up-to-and-including building for AWS’s Trainium architecture. And, in the long run, even the upside Anthropic scenario, where they have a compelling agent enterprise business, is compatible with AWS, which acts as infrastructure for many successful platform companies.
Third, AWS’s early investment in Bedrock was an early bet on AI optionality; the company’s investment in Trainium provides similar benefits in terms of the future of AI chips. You could certainly make the case that Amazon was and is behind in terms of core model offerings and chip infrastructure; that exact same case could be spun as Amazon being the best placed to shift as the future AI landscape becomes clearer over time.
AWS, meanwhile, remains the largest cloud provider by a significant margin, and, at the end of the day, enterprises would prefer to use the AI that is close to their existing data rather than to go through the trouble of a migration. And don’t forget about Alexa: there is, as I expected, precious little evidence of Alexa+ even being available, much less living up to its promise, but there is obviously more potential than ever in voice-controlled devices.
The Model Makers
The foundation model makers are obviously critical to AI; while this Article is focused on the Big Tech companies, it’s worth checking in on the status of the makers of the underlying technology:
OpenAI: I have long viewed OpenAI as the accidental consumer tech company; I think this lens is critical to understanding much of the last two years. A lot of OpenAI’s internal upheaval, for example, may in part stem from conflict with CEO Sam Altman, but it’s also the fact that early OpenAI employees signed up for a science project, not to be the next Facebook.
I do think that ChatGPT has won the consumer AI space, and is more likely to extend its dominance than to be supplanted. This, by extension, puts OpenAI fundamentally at conflict with any other entity that seeks to own the customer relationship, from Microsoft to Apple. Both companies, however, may have no choice but to make it work: Microsoft, because they’re already in too deep, and Apple, because OpenAI may, sooner rather than later, be the most compelling reason to buy the iPhone (if Apple continues to deepen its integration).
The big question in my mind is when and if OpenAI figures out an advertising model to supplement its subscription business. While I — and most of you reading this — will gladly pay for AI’s productivity gains, the fact remains that huge swathes of the consumer space likely won’t, and owning that segment not only locks out rivals, but also gives a long-term advantage in terms of revenue and the ability to invest in future models.
Anthropic: Anthropic may have missed out on the consumer space, but the company’s focus on coding has paid off in a very strong position with developers and a big API revenue stream. This is a riskier position in some respects, since developers and intermediaries like Cursor can easily switch to other models that might one day be better, but Anthropic is seeking to ameliorate that risk with products like Code that can not only be a business in its own right, but also generate critical data for improving the underlying model.
Anthropic’s reliance on Amazon and its Trainium chips for training is potentially suboptimal; it also could mean meaningful cost savings in the long run. Most importantly, however, Amazon is now a deeply committed partner for Anthropic; as I noted above, this is likely a much stabler situation than Microsoft and OpenAI.
Anthropic, unlike OpenAI, also benefits from its longer-term business opportunity being more aligned with the AGI dreams of its leading researchers: the latter might not create God, but if they manage to come up with an autonomous agent service along the way there will be a lot of money to be made.
xAI: I wrote more about xAI’s tenuous position last week; briefly:
xAI’s insistence on owning its own infrastructure seems to me to be more of a liability than an asset. Yes, Elon Musk can move more quickly on his own, but spending a lot of money on capital expenditures that aren’t fully utilized because of a lack of customers is an excellent way to lose astronomical amounts of money.
xAI is a company that everyone wants to exist as an alternative to keep OpenAI and Anthropic honest, but that doesn’t pay the bills. This is why the company should aggressively seek investment from Microsoft in particular.
There is an angle where xAI and Oracle make sense as partners: xAI could use an infrastructure partner, and Oracle could use a differentiated AI offering. The problem is that they could simply exacerbate each others challenges in terms of acquiring customers.
One of the most harmful things that has happened to xAI is the acquisition of X; that simply makes xAI a less attractive investment for most companies, and an impossible acquisition target for Meta, which is clearly willing to pay for xAI’s talent. What is more interesting is the relationship with Tesla; to the extent that the bitter lesson covers self-driving is the extent that xAI’s infrastructure can, at worst, simply be productively funneled to another Musk company.
Meta: Here we are full circle with the news of the week. There is a case to be made that Meta is simply wasting money on AI: the company doesn’t have a hyperscaler business, and benefits from AI all the same. Lots of ChatGPT-generated Studio Ghibli pictures, for example, were posted on Meta properties, to Meta’s benefit.
The problem for Meta — or anyone else who isn’t a model maker — is that the question of LLM-based AI’s ultimate capabilities is still subject to such fierce debate. Zuckerberg needs to hold out the promise of superIntelligence not only to attract talent, but because if such a goal is attainable then whoever can build it won’t want to share; if it turns out that LLM-based AIs are more along the lines of the microprocessor — essential empowering technology, but not a self-contained destroyer of worlds — then that would both be better for Meta’s business and also mean that they wouldn’t need to invest in building their own. Unfortunately for Zuckerberg, waiting-and-seeing means waiting-and-hoping, because if the bet is wrong Meta is MySpace.
This is also where it’s important to mention the elephant in the room: China. Much of the U.S. approach to China is predicated on the assumption that AI is that destroyer of worlds and therefore it’s worth risking U.S. technological dominance in chips to stop the country’s rise, but that is a view with serious logical flaws. What may end up happening — and DeepSeek pointed in this direction — is that China ends up commoditizing both chips and AI; if that happened it’s Big Tech that would benefit the most (to Nvidia’s detriment), and it wouldn’t be the first time.
If WWDC’s opening video — which cast Apple executives as characters in the upcoming F1 movie, with Senior Vice President of Software Engineering Craig Federighi in the starring role — was a bit of a fever dream, then the opening of Federighi’s presentation of Apple’s annual software updates had the air of a regretful admission the morning after that mistakes had been made.
To Federighi and Apple’s credit, there was no attempt to dance around the fact that last year’s WWDC ended up being a fever dream in its own right: Apple promised a litany of AI-enabled features, particularly for Siri, that have not shipped and may never ship. Federighi, after listing the basic and hardly ground-breaking Apple Intelligence features that did ship, admitted right up front:
As we’ve shared, we’re continuing our work to deliver the features that make Siri even more personal. This work needed more time to reach our high quality bar, and we look forward to sharing more about it in the coming year.
That’s only two sentences, of course, but the admission was notable and necessary; last year’s WWDC — which garnered high praise, including from yours truly — revealed that Something Is Rotten in the State of Cupertino. That was the title of John Gruber’s Daring Fireball article excoriating Apple for promising something it could not deliver:
Even with everything Apple overpromised (if not outright lied about) at the WWDC keynote, the initial takeaway from WWDC from the news media was wrongly focused on their partnership with OpenAI. The conventional wisdom coming out of the keynote was that Apple had just announced something called “Apple Intelligence” but it was powered by ChatGPT, when in fact, the story Apple told was that they — Apple — had built an entire system called Apple Intelligence, entirely powered by Apple’s own AI technology, and that it spanned from on-device execution all the way to a new Private Cloud Compute infrastructure they not only owned but are powering with their own custom-designed server hardware based on Apple Silicon chips. And that on top of all that, as a proverbial cherry on top, Apple also was adding an optional integration layer with ChatGPT.
So, yes, given that the news media gave credit for Apple’s own actual announced achievements to OpenAI, Apple surely would have been given even less credit had they not announced the “more personalized Siri” features. It’s easy to imagine someone in the executive ranks arguing “We need to show something that only Apple can do.” But it turns out they announced something Apple couldn’t do. And now they look so out of their depth, so in over their heads, that not only are they years behind the state-of-the-art in AI, but they don’t even know what they can ship or when. Their headline features from nine months ago not only haven’t shipped but still haven’t even been demonstrated, which I, for one, now presume means they can’t be demonstrated because they don’t work.
Gruber — my co-host on Dithering — has been writing and podcasting about Apple for over two decades; his podcast is called The Talk Show, and for the last ten years Apple executives have appeared for a live version of his show the week of WWDC. However, this year will be different; Gruber announced on Daring Fireball:1
Ever since I started doing these live shows from WWDC, I’ve kept the guest(s) secret, until showtime. I’m still doing that this year. But in recent years the guests have seemed a bit predictable: senior executives from Apple. This year I again extended my usual invitation to Apple, but, for the first time since 2015, they declined.
I think this will make for a fascinating show, but I want to set everyone’s expectations accordingly. I’m invigorated by this. See you at the show, I hope.
Neither Gruber nor (obviously) Apple said why the invitation was declined, but it was hard to not draw a line to Gruber’s viral article; Marco Arment said Apple was Retreating to Safety:
Maybe Apple has good reasons. Maybe not. We’ll see what their WWDC PR strategy looks like in a couple of weeks.
In the absence of any other information, it’s easy to assume that Apple no longer wants its executives to be interviewed in a human, unscripted, unedited context that may contain hard questions, and that Apple no longer feels it necessary to show their appreciation to our community and developers in this way.
I hope that’s either not the case, or it doesn’t stay the case for long.
This will be the first WWDC I’m not attending since 2009 (excluding the remote 2020 one, of course). Given my realizations about my relationship with Apple and how they view developers, I’ve decided that it’s best for me to take a break this year, gain some perspective, and decide what my future relationship should look like.
Maybe Apple’s leaders are doing that, too.
My biggest takeaway from WWDC is that Arment got it right: Apple is indeed “Retreating to Safety”. Retreating might, however, be exactly the right thing to do.
Apple’s Retreat to Its Core Competency
The headline feature of WWDC this year was Liquid Glass, a new unified design language that stretches across its operating systems. I will reserve judgment on Liquid Glass’s aesthetics and usability — Gruber likes it, but I am not one to install developer betas on my devices — but will make three meta observations.
First, hand-crafted UI overhauls are the polar opposite of the probabilistic world of generative AI. One is about deep consideration and iteration resulting in a finished product; the other is about in-the-moment token prediction resulting in an output that is ephemeral and disposable. Both are important and creative, but the downsides of that creativity — unfamiliarity and edge cases versus hallucination and false confidence — are themselves diametrically opposed.
Apple’s historical strengths have always been rooted in designing for finality; in my first year of Stratechery I did a SWOT analysis of the big tech companies and said about Apple:
Apple’s product development process is wonderful for developing finished products, but that same process doesn’t work nearly as well when it comes to building cloud services. Cloud services are never “done”; they are best developed by starting with a minimum viable product and then iterating based on usage. This is precisely opposite of what it takes to design a piece of hardware, and it’s a big reason why Apple struggles so much with cloud services (and why other services companies struggle with products). The canonical example of this, of course, was the MobileMe launch, which was delivered fully-formed and which, when faced with real world usage, crashed-and-burned. Apple’s latest offerings are better, but still suffer from too much internal development time per release. This is a hard problem to fix, because it touches the core of what makes Apple Apple.
I think it matters whether or not Liquid Glass is good, because it will be a testament about the state of Apple’s strengths; the point for this Article, however, is that WWDC was a retreat to those strengths, away from a technology that is very much inline with Apple’s historical weaknesses.2
Second, the core premise of the Liquid Glass re-design is leveraging Apple’s integration of hardware and software. This is how Vice President of Human Interface Design Alan Dye introduced Apple’s new design language:
Now with the powerful advances in our hardware, silicon, and graphics technologies, we have the opportunity to lay the foundation for the next chapter of our software. Today, we’re excited to announce our broadest design update ever. Our goal is a beautiful new design that brings joy and delight to every user experience, one that’s more personal and puts greater focus on your content. All while still feeling instantly familiar. And for the first time, we’re introducing a universal design across our platforms. This unified design language creates a more harmonious experience as he move between products, while maintaining the qualities that make each unique. Inspired by the physicality and richness of visionOS, we challenged ourselves to make make something purely digital, feel natural and alive, from how it looks to how it feels as it dynamically responds to touch.
To achieve this, we began by rethinking the fundamental elements that make up our software, and it starts with an entirely new expressive material we call Liquid Glass. With the optical qualities of glass, and a fluidity that only Apple can achieve, it transforms depending on your content, or even your context, and brings more clarity to navigation and controls. It beautifully refracts light, and dynamically reacts to your movement with specular highlights. This material brings a new level of vitality to every aspect of your experience, from the smallest elements you interact with, to larger ones, it responds in real time to your content, and your input, creating a more lively experience that we think you’ll find truly delightful. Elements once considered for rectangular displays have been redesigned to fit perfectly concentric with the rounded corners of the hardware. This establishes greater harmony between our software and hardware, while thoughtfully considered groups of controls, free up valuable space for your content. Liquid Glass is translucent and behaves just like glass in the real world. Its color is informed by your content and intelligently adapts between light and dark environments, and as a distinct functional layer that sits above your app, the material dynamically morphs when you need more options, or as you move between views.
We’ve always cared deeply about every detail of our software design, and it’s these moments of beauty, craft, and joy that bring our products to life. Our new design blurs the lines between hardware and software to create an experience that’s more more delightful than ever, while still familiar and easy to use. Today marks an exciting and beautiful new chapter for our design, one that sets the stage for our next era of our products and how you interact with them every day.
Sebastiaan De With, in a remarkably prescient post predicting the nature of this new design, emphasized how only Apple could make Liquid Glass:
A logical next step could be extending physicality to the entirety of the interface. We do not have to go overboard in such treatments, but we can now have the interface inhabit a sense of tactile realism. Philosophically, if I was Apple, I’d describe this as finally having an interface that matches the beautiful material properties of its devices. All the surfaces of your devices have glass screens. This brings an interface of a matching material, giving the user a feeling of the glass itself coming alive…
The interfaces of computers of the future are often surprisingly easy to imagine. We often think of them and feature them in fiction ahead of their existence: our iPhone resembles a modern Star Trek tricorder; many modern AI applications resemble the devices in sci-fi movies like ‘Her’ and (depressingly) Blade Runner 2049. It’s not surprising, then, that concept interfaces from the likes of Microsoft often feature ‘glass fiction’:
The actual interface is unfortunately not nearly as inspired with such life and behavioral qualities. The reason is simple: not only is the cool living glass of the video way over the top in some places, but few companies can actually dedicate significant effort towards creating a hardware-to-software integrated rendering pipeline to enable such UI innovations…Only Apple could integrate sub pixel antialiasing and never-interrupted animations on a hardware level to enable the Dynamic Island and gestural multi-tasking; only Apple can integrate two operating systems on two chips on Vision Pro so they can composite the dynamic materials of the VisionOS UI. And, perhaps only Apple can push the state of the art to a new interface that brings the glass of your screen to life.
De With’s prescience actually gives me great hope for Liquid Glass: the best innovations are obvious to those who understand what is just becoming possible, and Apple’s integration has been and continues to be a meaningful advantage for things like user interfaces.
Third, Apple CEO Tim Cook has for a long time extended his framing of Apple’s differentiation to be the integration of hardware, software, and services, but while that is certainly true from a financial perspective, I’ve long had a hard time buying that the services component made for better products; as I noted above, Apple’s services have typically been something to be endured, as opposed to a reason to buy their devices, and the Siri debacle has only emphasized that point.
What is much more compelling — and the fact that Liquid Glass is a design language meant to unify Apple’s devices speaks to this — is the integration of Apple’s devices with each other. Every Apple product you buy is enhanced by the purchase of additional Apple products; to that end, one of the coolest parts of the WWDC presentation was about Continuity, Apple’s brand for features that connect various Apple products:
Let’s talk about the power of continuity. Continuity helps you work seamlessly across Apple devices, and we’re excited to introduce two new Continuity features. First, we’re bringing Live Activities to Mac. So if you’ve ordered Uber Eats on your iPhone, the Live Activity also appears in the menu bar, and when you click, the app opens in iPhone-mirroring, so you can take action directly on your Mac. We’re also enhancing the calling experience by bringing the Phone app to Mac. You can conveniently access your familiar content, like recents, contacts, and voicemail, synced from iPhone, and easily make a call with just a click. Incoming calls look beautiful on the bigger screen, featuring stunning contact posters of your friends and family, and the Phone app on Mac includes all the great updates we talked about earlier, like hold assist, call screening, and live translation. So that’s what’s new in Continuity.
These sorts of features aren’t going to change the world; they are, though, features that I can absolutely see making my life better and more convenient on an ongoing basis. And, to the broader point, they are features that only Apple can do.
More generally, yes, a focus on UI design is a retreat from The Gen AI Bridge to the Future; that future, however, will start from the devices we still use all day every day, and Apple focusing on making those devices better is a retreat that I expect will have a far more positive impact on my life than the company struggling to catch up in AI.
Apple’s Retreat to Empowering Developers and Partners
That’s not to say there weren’t some notable AI announcements in Apple’s keynote.
First, Apple announced the Foundation Models framework:
This year we’re doing something new, and we think it’s going to be pretty big. We’re opening up access for any app to tap directly into the on-device, large language model at the core of Apple Intelligence, with a new Foundation Models Framework. This gives developers direct access to intelligence that’s powerful, fast, built with privacy, and available even when you’re offline. We think this will ignite a whole new wave of intelligent experiences in the apps you use every day.
For example, if you’re getting ready for an exam, an app like Kahoot can create a personalized quiz from your notes to make studying more engaging, and because it uses on-device models, this happens without Cloud API costs. Or perhaps you’re camping off-grid, poring over the hikes you downloaded to AllTrails. Just describe what you’re in the mood for, and AllTrails can use our on-device models to suggest the best option. We couldn’t be more excited about how developers can build on Apple Intelligence to bring you new experiences that are smart, available when you’re offline, and that protect your privacy.
It’s important not to oversell the capabilities of Apple’s on-device AI models: of course developers who want to create something that is competitive with the output of something like ChatGPT will need to use cloud-based AI APIs. That reality, however, applies to Apple as well! Part of the folly of the initial Apple Intelligence approach is that Apple was promising to deliver beyond state-of-the-art capabilities on the cheap, using its users’ processors and power.
What is compelling about the Foundation Models Framework is how it empowers small developers to experiment with on-device AI for free: an app that wouldn’t have AI at all for cost reasons now can, and if that output isn’t competitive with cloud AI then that’s the developer’s problem, not Apple’s; at the same time, by enabling developers to experiment Apple is the big beneficiary of those that discover how to do something that is only possible if you have an Apple device.
Second, Apple deepened its reliance on OpenAI, incorporating ChatGPT’s image generation capabilities into Image Playground and adding ChatGPT analysis to Visual Intelligence. There is still no sign of the long-rumored Gemini integration or the ability to switch out ChatGPT for the AI provider of your choice, but the general trend towards relying on partners who are actually good at building AI is a smart move.
Third, Apple is also incorporating ChatGPT much more deeply into Xcode, its Integrated Development Environment (IDE) for building apps for Apple platforms; developers can also plug in other models using API keys. Xcode still has a long ways to go to catch up to AI-first IDEs like Cursor, but again, partnering with foundational model makers is a far smarter strategy than Apple trying to do everything itself.
These are, to be sure, obviousmoves, but that doesn’t make them any less important, both in terms of Apple’s future, and also with regard to the theme of this Article: Apple’s initial success with the Apple II was because of 3rd-party developers, and developers were critical to making the iPhone a sustainable success. Trusting developers and relying on partners may be a retreat from Apple’s increasing insistence on doing everything itself, but it is very much a welcome one.
Apple’s [Forced] Retreat to App Store Sanity
Apple didn’t say much about the App Store in the keynote, but they did announce a new Games app; M.G. Siegler theorized late last month that this may be laying the groundwork for splitting up Games from the rest of the App Store:
What if this new gaming-focused app – let’s just call it ‘Game Store’ – is not only meant to unify Apple’s gaming-focused efforts, but also to separate them from the App Store itself? Why might Apple do this? Because this would allow them to more easily differentiate between the two and, importantly, give the two independent policies.
That means that Apple could, say, drop the rate developers have to pay when it comes to revenue share in the App Store, while keeping it the same as it is now in the ‘Game Store’. And that matters because actually, gaming makes up some two-thirds of Apple’s App Store revenue at the moment (between paid downloads and in-app purchases – but it’s predominantly the latter). It’s the actual key to Apple’s model for this segment of the Services business.
And guess what else is true? In gaming, a 70/30 split is a well-established norm. In fact, it’s where Apple’s own App Store split originates from (by way of iTunes, which also copied the model back in the day)! Yes, there are others who have tried to disrupt this split, notably Epic, but Apple has a much stronger case for a 70/30 split when it comes to gaming than it now does for the overall app ecosystem.
So hear me out: the ‘Game Store’ keeps the 70/30 model and the ‘App Store’ moves to something more like 85/15 as the standard (matching Apple’s currently convoluted system for small developers with various arbitrary thresholds). Perhaps for smaller developers, Apple even drops it to 90/10.
Apple did not announce such a shift yesterday, but the infrastructure is now in place to do exactly what I have advocated for years: treat games differently from other apps. Gaming revenue is almost entirely based on zero marginal cost content, and games are more susceptible to abuse and more likely to be used by kids; I don’t mind Apple’s more heavy-handed approach in that case and, as Siegler notes, this level of control is the industry standard for other gaming platforms like consoles. In other words, Apple should retreat from trying to take a cut of everything digital, and instead act like a console maker where appropriate, and a more neutral computer platform for everything else.
Unfortunately for Apple, keeping console-level control of games may no longer be possible, particularly after the Ninth Circuit Court of Appeals denied Apple’s request for a stay of Judge Yvonne Gonzalez Rogers’ order lifting anti-steering restrictions on all apps, including games. The functional outcome of Gonzalez Rogers’ order is a retreat by Apple from its overbearing control of app monetization, albeit not one Apple is engaged in willingly.
Once again, however, a retreat is exactly what Apple needs. The company has gone too far with the App Store, not only embittering developers and losing court cases, but also has put its fundamental differentiation at risk. I warned the company of exactly this in 2021’s Integrated Apple and App Store Risk:
This is where the nuance I discussed in App Store Arguments becomes much more black-and-white. Yes, Apple created the iPhone and the App Store and, under current U.S. antitrust doctrine, almost certainly has the right to impose whatever taxes it wishes on third parties, including 30% on purchases and the first year of subscriptions, and completely cutting off developers from their customers. Antitrust law, though, while governed by Supreme Court precedent, is not a matter of constitutionality: it stems from laws passed by Congress, and it can be changed by new laws passed by Congress.
One of the central planks of many of those pushing for new laws in this area are significant limitations on the ability of platforms to offer apps and services, or integrate them in any way that advantages their offerings. In this potential world it’s not simply problematic that Apple charges Spotify 30%, or else forces the music streaming service to hope that users figure out how to subscribe on the web, even as Apple Music has a fully integrated sign-up flow and no 30% tax; it is also illegal to incorporate Apple Music into SharePlay or Shared-with-you or Photos, or in the most extreme versions of these proposed laws, even have Apple Music at all. This limitation would apply to basically every WWDC announcement: say good-bye to Quick Note or SharePlay-as-an-exclusive-service, or any number of Apple’s integrated offerings.
I think these sorts of limitations would be disappointing as a user — integration really does often lead to better outcomes sooner — and would be a disaster for Apple. The entire company’s differentiation is predicated on integration, including its ability to abuse its App Store position, and it would be a huge misstep if the inability to resist the latter imperiled the former.
This, more than anything, is why Apple should rethink its approach to the App Store. The deeper the company integrates, the more unfair its arbitrary limits on competing services will be. Isn’t it enough that Spotify will never be as integrated as Apple Music, or that 1Password will not be built-in like Keychain, or that SimpleNote will only ever be in its sandbox while Apple Notes is omnipresent? Apple, by virtue of building the underlying platform, has every advantage in the world when it comes to offering additional apps and services, and the company at its best leverages that advantage to create experiences that users love; in this view demanding 30% and total control of the users of its already diminished competition isn’t simply anticompetitive, it is risking what makes the company unique.
This is exactly what is happening in Europe: the DMA requires Apple to open up a whole host of capabilities that undergird its integration both on individual devices and especially between devices, and Apple is signalling that it will simply remove those features in the E.U.. That is one way to solve the company’s DMA issues, but the cost is severe: in one of Apple’s largest markets it can’t actually deliver on the differentiation that, earlier in this Article, I was celebrating its retreat to.
Retreat and Reset
The biggest part of WWDC — and the biggest part of this Article — was about Liquid Glass, which has drawn some unflattering comparisons to past Microsoft operating system releases:
The Windows Vista update with aero glass was a huge part of my childhood, so I'm getting serious flashbacks
Again, I’ll withhold judgment on Liquid Glass until it ships, but there is another Microsoft OS comparison I’ve been thinking about recently: to me last year’s Siri disaster is a lot like Windows 8.
Windows 8 was an attempt to leverage Microsoft’s large PC install base into a competitive position in touch-based devices, and it did not go well: consumers — particularly enterprises — hated the new UI, and developers weren’t interested in a platform without users. Microsoft was forced to retreat, and eventually came out with Windows 10, which was much more inline with traditional Windows releases.
Apple has clearly missed the boat on cutting edge AI; what I’m open to is the argument that this was a ship the company was never meant to board, at least when it comes to products like ChatGPT. Meanwhile, I’ve long been convinced that Apple has gone too far in its attempt to control everything even tangentially related to its devices; from 2017’s Apple and the Oak Tree:
Apple has had a special run, thanks to its special ability to start with the user experience and build from there. It is why the company is dominant and will continue to be so for many years. Apple as an entity, though, is not special when it comes to the burden of success: there will be no going back to “Rip. Mix. Burn.” or its modern equivalent.
In short, Apple is no longer the little reed they were when Jobs could completely change the company’s strategy in mere months; the push towards ever more lock-in, ever more centralization, ever more ongoing monetization of its users — even at the cost of the user experience, if need be — will be impossible to stop, for Tim Cook or anyone else. After all, such moves make Apple strong, until of course they don’t.
To that end, while I understand why many people were underwhelmed by this WWDC, particularly in comparison to the AI extravaganza that was Google I/O, I think it was one of the more encouraging Apple keynotes in a long time. Apple is a company that went too far in too many areas, and needed to retreat. Focusing on things only Apple can do is a good thing; empowering developers and depending on partners is a good thing; giving even the appearance of thoughtful thinking with regards to the App Store (it’s a low bar!) is a good thing. Of course we want and are excited by tech companies promising the future; what is a prerequisite is delivering in the present, and it’s a sign of progress that Apple retreated to nothing more than that.
I have come to believe that advertising is the original sin of the web. The fallen state of our Internet is a direct, if unintentional, consequence of choosing advertising as the default model to support online content and services. Through successive rounds of innovation and investor storytime, we’ve trained Internet users to expect that everything they say and do online will be aggregated into profiles (which they cannot review, challenge, or change) that shape both what ads and what content they see. Outrage over experimental manipulation of these profiles by social networks and dating companies has led to heated debates amongst the technologically savvy, but hasn’t shrunk the user bases of these services, as users now accept that this sort of manipulation is an integral part of the online experience.
Marc Andreessen, who was there when the web was born, explained in a 2019 podcast why this sin was committed (this quote is lightly edited for clarity):
One would think the most obvious thing to do would be building in the browser the ability to actually spend money, right? You’ll notice that didn’t happen, and in a lot of ways, we don’t even think it’s unusual that that didn’t happen, because maybe that shouldn’t have happened. I think the original sin was we couldn’t actually build economics, which is to say money, into the core of the internet and so therefore advertising became the primary business model…
We tried very hard to build payments into the browser. It was not possible…We made a huge mistake. We tried to work with the banks and we tried to work with the credit card companies…it was sort of the classic kind of single point of failure bottleneck, or at least in this case, two points of failure. Visa and MasterCard essentially had a duopoly at the time, and so they were just literally, if they did not want you to be in the switch, they did not want you to be able to do transactions, you just simply weren’t going to do it.
I think Andreessen is too hard on himself, and I think Zuckerman is too harsh on the model Andreessen created the conditions for. The original web was the human web, and advertising was and is one of the best possible ways to monetize the only scarce resource in digital: human attention. The incentives all align:
Users get to access a vastly larger amount of content and services because they are free.
Content makers get to reach the largest possible audience because access is free.
Advertisers have the opportunity to find customers they would have been never able to reach otherwise.
Yes, there are the downsides to advertising Zuckerman fretted about, but everything is a trade-off, and the particular set of trade-offs that led to the advertising-centric web were, on balance, a win-win-win that generated an astronomical amount of economic value.
Moreover, I disagree with Andreessen that we could have ended up with a better system if the banks and credit card companies had been willing to play ball. In fact, over the last thirty years, the credit card companies in particular have — in part thanks to companies like Stripe — gotten their digital acts together, and are integral to a huge amount of web-based commerce, which itself is driven through digital advertising (the largest category of advertising for both Google and Meta). That too is human, in that the biggest outcome of digital advertising is physical products and real-world experiences like travel (digital products like apps and games, meanwhile, are themselves pursuing human attention).
What was not viable in the 1990s, nor at any time since then, was something like micro-transactions for content. One obvious problem is that the fee structure of credit cards don’t allow for very small transactions; another problem is that the costs to product content are front-loaded, and the potential payoff is both back-loaded and unpredictable, making it impossible to make a living. The biggest problem of all, however, is that micro-transactions are anti-human: forcing a potential content consumer to continually decide on whether or not to pay for a piece of content is alienating, particularly when plenty of alternatives for their scarce attention exist.
Subscriptions do work at smaller scales, because they are ultimately not about paying for content, but giving money to another human (or human institution); from The Local News Business Model:
It is very important to clearly define what a subscriptions means. First, it’s not a donation: it is asking a customer to pay money for a product. What, then, is the product? It is not, in fact, any one article (a point that is missed by the misguided focus on micro-transactions). Rather, a subscriber is paying for the regular delivery of well-defined value.
Each of those words is meaningful:
Paying: A subscription is an ongoing commitment to the production of content, not a one-off payment for one piece of content that catches the eye.
Regular Delivery: A subscriber does not need to depend on the random discovery of content; said content can be delivered to the subscriber directly, whether that be email, a bookmark, or an app.
Well-defined Value: A subscriber needs to know what they are paying for, and it needs to be worth it.
This last point is at the crux of why many ad-based newspapers will find it all but impossible to switch to a real subscription business model. When asking people to pay, quality matters far more than quantity, and the ratio matters: a publication with 1 valuable article a day about a well-defined topic will more easily earn subscriptions than one with 3 valuable articles and 20 worthless ones covering a variety of subjects. Yet all too many local newspapers, built for an ad-based business model that calls for daily content to wrap around ads, spend their limited resources churning out daily filler even though those ads no longer exist.
I expect that this model will endure in the age of AI; obviously I’m biased on this point, but in a world of infinite content-on-demand, common content becomes community: if I’m successful this essay will generate a lot of discussion amongst a lot of people precisely because it is both original and widely accessible, funded by an audience that wants me to keep on writing Articles exactly like this.
The Death of the Ad-Supported Web
The ad-supported web — particularly text-based sites — is going to fare considerably worse. In fact, the most substantive pushback to my defense of advertising was in my own excerpt: most ad-supported content is already terrible, thanks to the bad incentives both Zuckerman and Andreessen bemoaned, and the impossible economics enabled by zero marginal cost content generation and consumption.
The center of this world for the last twenty years has been Google.
Google in its most idealized form Aggregated content consumers by mastering discovery in this world of abundance, directing users to exactly the site they were looking for, which was monetized through ads that were sold and served by Google. Indeed, this is the great irony in the ads antitrust case in which Google is currently embroiled; Eric Seufert asked on MobileDevMemo:
I’ve heard arguments that, because Google suppressed competition in open web advertising markets, those markets should flourish when Google’s monopoly is broken. But my sense is that this ignores two realities. First, that consumer engagement has shifted into apps and walled gardens irreversibly. And second, that Google was keeping the open web on life support, and the open web’s demise will be hastened when Google no longer has an incentive to support it. What happens to the open web when its biggest, albeit imperfect, benefactor loses the motivation to sustain it?
Note Seufert’s two points: walled gardens like social networks are both more attractive to most users and also better for advertisers, and Google might soon lose what little motivation they had left to support the open web. However, that’s not Google’s — and the web’s — only problem. Why go through the hassle of typing a search term and choosing the best link — particularly as search results are polluted by an increasingly overwhelming amount of SEO spam, now augmented by generative AI — when ChatGPT (or Google itself) will simply give you the answer you are looking for?
In short, every leg of the stool that supported the open web is at best wobbly: users are less likely to go to ad-supported content-based websites, even as the long tail of advertisers might soon lose their conduit to place ads on those websites, leaving said websites even less viable than they are today — and they’re barely hanging on as it is!
Microsoft and the Open Agentic Web
This reality is the fly in the ointment of an intriguing set of proposals that Microsoft put forward yesterday at the Build 2025 Developer Conference about “The Open Agentic Web”; here’s CTO Kevin Scott:
The thing that is super important if you think about what an open-agentic web could be, is you need agents to be able to take actions on your behalf, and one of the really important things about agents being able to take actions on your behalf is they have to be plumbed up to the greater world. So you need protocols, things like MCP and A2A and things that likely are going to be emerging over the coming year that will help connect in an open, reliable, interoperable way the agents that you are writing and agents that are being used so actively now by hundreds of millions of people to be able to go access content, to access services, to take action on behalf of users in fulfilling the tasks that have been delegated to them.
One aspect of this vision of the agentic web was Microsoft’s commitment to the Model Context Protocol created by Anthropic; Scott told Nilay Patel in an excellent interview in The Verge that while MCP wasn’t exactly what he would have designed from scratch, ubiquity is more important than semantic differences, particularly when you’re trying to create HTTP for AI agents.
The second part of Scott’s vision was something Microsoft created called NLWeb, a natural language interface for websites that makes them more directly accessible for agents:
If you think back to the web, we have HTTP, and then we had things that sit on top of HTTP, like HTML mainly, that are opinionated about the payload, and so we’re announcing today NLWeb. The idea behind NLWeb is it is a way for anyone who has a website or an API already to very easily make their website or their API an agentic application. It lets you implement and leverage the full power of large language models to enrich the services and products that you’re already offering, and because every NLWeb endpoint is by default an MCP server, It means that those things that people are offering up via NLWeb will be accessible to any agent that speaks MCP. So you really can think about it a little bit like HTML for the agentic web.
We have done a bunch of work already with partners who have been really excited and been able to really very quickly get to quick implementations and prototypes using NLWeb. We’ve worked with TripAdvisor, O’Reilly Media, a ton of really great companies that offer important products and services on the internet to add in a web functionality to their sites so that they can have these agentic experiences directly on their sites.
Scott concluded by re-emphasizing how important it was that the layers of the agentic web be open, and used the evolution of the Internet as his example of why:
So the last thing that I want to say before handing things back over to Satya is to just sort of press on these two points about why like open is so important here. So you know it is unbelievable what can happen in the world when simple components and simple protocols that are composable with one another are out there, exposed to the full scrutiny and creativity of every developer in the world who wants to participate or who has an idea.
This thought game that I play with myself all the time is trying to imagine what the web would have looked like if one of the actors in the early development of the web, say the browser manufacturers, had decided that they wanted to vertically integrate and own the entire web. A hundred percent of the web would have been…dictated by the limits of their imagination, and it’s just obvious with 30 years of history that that wouldn’t have been a very interesting web. The web is interesting because millions, tens of millions, hundreds of millions of people are participating to make it into this rich dynamic thing. That’s what we think we need with the agentic web, and that’s what we’re hoping you all can get inspired to go work on a little bit, to riff on, and to use the full extent of your imagination to help make this thing interesting.
I think that widespread adoption of MCP as a protocol layer and NLWeb as a markup layer sounds excellent; the big hole in Scott’s proposal, however, was pointed out by Patel in that interview:
That is the piece that on the web right now seems most under threat, the underlying business dynamics of I start a website, I put in a bunch of schema that allows search engines to read my website and surface my content across different distributions. I might add an RSS feed, which is a standardized distribution that everyone uses and agrees on. There’s lots of ways to do this.
But if I make a website, I open myself up to distribution on different surfaces. What I will get in return for that is not necessarily money — almost in every case, not money. What I’ll get is visitors to my website, and then I’ll monetize them however I choose to: selling a subscription, display ads, whatever it is. That’s broken, right? As more and more of the answers appear directly, particularly in AI-based search products, traffic to websites has generally dropped. We see this over and over again. What’s going to replace that in the agentic era, where we’ve created new schema for agents to come and talk to my website and receive some answers? What’s going to make that worth it?
Scott in his answer noted that websites would be able to communicate to agents what they wanted to make available and on what terms, along with some vague hand-waving about new advertising models and transactions. The last point is valid: Trip Advisor sells hotel rooms, and O’Reilly sells training courses, and you can see a world where websites based on transactions can not only benefit from exposing themselves to agents, but in fact transact more (and potentially pay an affiliate fee). Patel, however, rightly pressed Scott on the prospects for ad-supported content sites:
As Google keeps more of the traffic for itself or it thinks differently about training data, all this stuff is changing. The trade here is make your website more agentic, and then MCP as a protocol will allow you to build some new business models on it. The problem, as I see it, is that the traffic to the web is in precipitous decline as Google referrals go into decline. How do you fix that problem so that everyone is incentivized to keep building on the web?
I don’t know, honestly.
“The Original Sin” of the Internet lacking native payments was not, in my opinion, a sin at all: advertising supported the human web not because Andreessen failed to make a deal with the credit card companies, but because it was the only business model that made sense.
No, the real neglect and missed opportunity in terms of payments is happening right now: Microsoft is on to the right idea with its adoption of MCP and introduction of NLWeb, but its proposal, by virtue of not including native payments, isn’t nearly as compelling as it should be. The key difference from the 1990s is that on the agentic web native digital payments are both viable and the best possible way to not only keep the web alive, but also in the process create better and more useful AI.
Stablecoin legislation overcame a procedural blockade in the US Senate, marking a major victory for the crypto industry after a group of Democrats dropped their opposition Monday. The industry-backed regulatory bill is now set for debate on the Senate floor with a bipartisan group hoping to pass it as soon as this week, although senators said a final vote could slip until after the Memorial Day recess.
I know I have driven long-time Stratechery readers a bit batty with my long-running and still-enduring-in-the-face-of-massive-grift-and-seemingly-unending-scandals interest in crypto, but stablecoins are genuinely a big deal. I wrote a brief explainer last fall when Stripe acquired Bridge:
Stablecoins distill crypto to these most interesting bits. Unlike Bitcoin, stablecoins do not have intrinsic value downstream from a network effect, and unlike Ethereum, they are not set up to execute smart contracts or other applications; rather, their value is right there in the name: they are stable representations of value — usually the U.S. dollar…What remains is a synthetic currency that is digital but scarce, with all of the affordances that allows for, including the ability to move money frictionlessly (thus Collison’s analogy). The analogy I think of is to the Internet itself:
Physical goods are scarce, but while you can scale up from hand-delivery, you still have to pay a lot for a delivery service, and if you cross borders you have to deal with customs.
Information used to be delivered in person, then via physical media like letters or newspapers, but now it is purely digital and freely distributed and duplicated all over the world; it is abundant.
Dollars right now are more akin to physical goods than they are to information: you can deliver it by hand, or even via delivery services like ACH or SWIFT, but the need for verification and confirmation flows introduce a ton of friction. Moreover, you can’t actually do anything with dollars at rest, other than watch them deflate. Stablecoins solve these problems: you can transfer them like information, while preserving scarcity, while blockchains provide verification and confirmation that scales from the smallest transactions to the biggest; meanwhile, they also earn a return while at rest thanks to the assets backing them.
Stablecoins solve several of the micro-transaction problems I listed above, including dramatically lower — or no — fees, and the fact that they are infinitely divisible, and thus can scale to very small amounts. Stablecoins, by virtue of being programmable, are also well-suited to agents; agents, meanwhile, are much more suited to micro-transactions, because they are, in the end, simply software making a decision, unencumbered by the very human feeling of decision paralysis.
In fact, we already have an excellent example of (deterministic) agents making micro-transactions at scale: the entire digital ads ecosystem! Every time a human loads a webpage, an awe-inspiring amount of computation and communication happens in milliseconds, as an auction is run to fill the inventory on that page with an ad that is likely to appeal to the human. These micro-transactions are only worth fractions of a penny, but the aggregate volume of them drives trillions of dollars worth of value.
The problem, as both I and Patel noted, is that this ecosystem depends on humans seeing those webpages, not impersonal agents impervious to advertising, which destroys the economics of ad-supported content sites, which, in the long run, dries up the supply of new content for AI. OpenAI and Google in particular are clumsily addressing the supply issue by cutting deals with news providers and user-generated content sites like Reddit; this, however, is bad for the sort of competition Microsoft wants to engender, and ultimately won’t scale to the amount of new content that needs to be generated.
What is possible — not probable, but at least possible — is to in the long run build an entirely new marketplace for content that results in a new win-win-win equilibrium.
First, the protocol layer should have a mechanism for payments via digital currency, i.e. stablecoins. Second, AI providers like ChatGPT should build an auction mechanism that pays out content sources based on the frequency with which they are cited in AI answers. The result would be a new universe of creators who will be incentivized to produce high quality content that is more likely to be useful to AI, competing in a marketplace a la the open web; indeed, this would be the new open web, but one that operates at even greater scale than the current web given the fact that human attention is a scarce resource, while the number of potential agents is infinite.
There is, to be sure, a tremendous amount of complexity in what I am proposing, and the path to a marketplace for data generation is quite unclear at the moment. Who, however, could have predicted exactly how the ad-supported web would have evolved, or centrally designed the incredible complexity that undergirds it?
This is where Scott’s exhortation of openness is spot on: a world of one dominant AI making business development deals with a few blessed content creators, and scraping the carcass of what remains on the web for everything else, is a far less interesting one than one driven by marketplaces, auctions, and aligned incentives.
To get there, however, means realizing that the Internet’s so-called “Original Sin” was in fact the key to realizing the human web’s potential, while the actual mistake would be in not building in payments now for the coming agentic web.
I cannot accept your canon that we are to judge Pope and King unlike other men, with a favourable presumption that they did no wrong. If there is any presumption it is the other way against holders of power, increasing as the power increases. Historic responsibility has to make up for the want of legal responsibility. Power tends to corrupt and absolute power corrupts absolutely. — Lord Acton, Letter to Bishop Creighton
The first thing you might notice is that while the Update was from 2015, I had the wrong year in my email; I guess the positive spin is that that was due to the duct-tape-and-wire nature of my publishing system back then, but it was an embarrassing enough error that I never did link to Salmon’s piece. The reason I mention it now, however, is that while Salmon had positive thing to say about my coverage of net neutrality and Microsoft’s then-new Outlook app, he was mostly bemused by my coverage of the App Store:
Thompson is proud to have obsessions, and one of his geeky obsessions is the arcane set of rules surrounding apps in Apple’s app store. His conclusion is also a way of continuing a thread which runs through many past and future updates: as such it’s a way of rewarding loyal readers.
I am not writing this Article, however, to say “I told you so”; rather, what strikes me about my takes at the time, including the one that Salmon highlighted, is what I got wrong, and how much the nature of my errors bums me out.
Apple Power
The anticompetitive nature of Apple’s approach to the App Store revealed itself very early; John Gruber was writing about The App Store’s Exclusionary Policies just months after the App Store’s 2008 launch. The prompt was Apple’s decision to not approve an early podcasting app because “it duplicate[d] the functionality of the Podcast section of iTunes”; Gruber fretted:
The App Store concept has trade-offs. There are pros and cons to this model versus the wide-open nature of Mac OS X. There are reasonable arguments to be made on both sides. But blatantly anti-competitive exclusion of apps that compete with Apple’s own? There is no trade-off here. No one benefits from such a policy, not even Apple. If this is truly Apple’s policy, it’s a disaster for the platform. And if it’s not Apple’s policy, then Podcaster’s exclusion is proof that the approval process is completely broken.
Apple eventually started allowing podcast apps a year or so later (without any formal announcement), but the truth is that there wasn’t any evidence that Apple was facing any sort of disaster for the platform. Gruber himself recognized this reality two years later in an Article about Adobe’s unhappiness with the App Store:
It’s folly to pretend there aren’t trade-offs involved — that for however much is lost, squashed by Apple’s control, that different things have not been gained. Apple’s control over the App Store gives it competitive advantages. Users have a system where they can install apps with zero worries about misconfiguration or somehow doing something wrong. That Adobe and other developers benefit least from this new scenario is not Apple’s concern. Apple first, users second, developers last — those are Apple’s priorities.
Gruber has returned to this point about Apple’s priority stack regularly over the years, even as some of the company’s more egregious App Store policies seemed to benefit no one but Apple itself. Who benefits from needing to go to Amazon in the browser to buy Kindle books, or there being no “subscription” option in Netflix?1 Judge Yvonne Gonzalez Rogers argued this lack of user consideration extended to Apple’s anti-steering provision, which forbade developers from telling users about better offers on their websites, and linking to them; from Gonzalez Rogers’ original opinion:
Looking at the combination of the challenged restrictions and Apple’s justifications, and lack thereof, the Court finds that common threads run through Apple’s practices which unreasonably restrains competition and harm consumers, namely the lack of information and transparency about policies which effect consumers’ ability to find cheaper prices, increased customer service, and options regarding their purchases. Apple employs these policies so that it can extract supracompetitive commissions from this highly lucrative gaming industry. While the evidence remains thin as to other developers, the conclusion can likely be extended.
More specifically, by employing anti-steering provisions, consumers do not know what developers may be offering on their websites, including lower prices. Apple argues that consumers can provide emails to developers. However, there is no indication that consumers know that the developer does not already have the email or what the benefits are if the email was provided. For instance, Apple does not disclose that it serves as the sole source of communication for topics like refunds and other product-related issues and that direct registration through the web would also mean direct communication. Consumers do not know that if they subscribe to their favorite newspaper on the web, all the proceeds go to the newspaper, rather than the reduced amount by subscribing on the iOS device.
While some consumers may want the benefits Apple offers (e.g., one-stop shopping, centralization of and easy access to all purchases, increased security due to centralized billing), Apple actively denies them the choice. These restrictions are also distinctly different from the brick-and-mortar situations. Apple created an innovative platform but it did not disclose its rules to the average consumer. Apple has used this lack of knowledge to exploit its position. Thus, loosening the restrictions will increase competition as it will force Apple to compete on the benefits of its centralized model or it will have to change its monetization model in a way that is actually tied to the value of its intellectual property.
This all seems plausible, and, thanks to Judge Gonzalez Rogers’ latest ruling, is set to be tested in a major way: apps like Spotify have already been updated to inform users about offers on their websites, complete with external links. Moreover, those links don’t have to follow Apple’s proposed link entitlement rules, which means they can be tokenized to the app user, facilitating a fairly seamless checkout experience, without the need to login separately.
Still, there are strong arguments to be made that many apps may be disappointed in their web purchase experience; Apple’s in-app purchase flow is so seamless and integrated — and critically, linked to an up-to-date payment method — that it will almost certainly convert better than web-based flows. At the same time, a 30% margin difference is a strong incentive to close that gap; the 15% margin difference for subscriptions is smaller, but at the same time, the payoffs from a web-based subscription — no Apple tax for the entire lifetime of the user — are so significant that the incentives might even be stronger.
Those incentives are likely to accrue to users: Spotify could, for example, experiment with offering some number of months free, or lower prices for a year, or just straight up lower prices overall; this is in addition to the ability to offer obvious products that have previously been impossible, like individual e-books. This is good for users, and it’s good for Spotify.
What is notable — and what I got wrong all those years ago — is the extent to which this is an unequivocally bad thing for Apple. They will, most obviously, earn less App Store revenue than they might have otherwise; while not every purchase on the web is one not made in the App Store — see previously impossible products, like individual e-books — the vast majority of web-based revenue earned by app makers will be a direct substitute for revenue Apple previously took a 15–30% cut of. Apple could, of course, lower their take rate, but that makes the point!
At the same time, I highly doubt that web-based purchases will lead to any increase in Apple selling more iPhones (they might, however, sell more advertising). This is the inverse of Gruber’s long-ago concern about Apple’s policies being “a disaster for the platform”, or my insistence that the company’s policies were “unsustainable”. In fact, they were quite sustainable, and extremely profitable.
The Chicken-and-Egg Problem
Before I started Stratechery, I worked at Microsoft recruiting developers for the Windows App Store; the discussion then was about the “chicken-and-egg problem” of building out a new platform: to get users you needed apps, but to get developers to build those apps you needed users that they wished to reach. Microsoft tried to cold-start this conundrum on the only side where they had a hope of exerting influence, which was developers: this meant lots of incentives for app makers, up-to-and-including straight up paying them to build for the platform.
This made no difference at all: most developers said no, cognizant that the true cost of building an app for a new platform was the ongoing maintenance of said app for a limited number of people, and those that said yes put forth minimal effort. Even if they had built the world’s greatest apps, however, I don’t think it would have mattered.
The reality is that platforms are not chicken-and-egg problems: it is very clear what comes first, and that is users. Once there are users there is demand for applications, and that is the only thing that incentivizes developers to build. Moreover, that incentive is so strong that it really doesn’t matter how many obstacles need to be overcome to reach those users: that is why Apple’s longstanding App Store policies, egregious though they may have been, ultimately did nothing to prevent the iPhone from having a full complement of apps, and, by extension, did nothing to diminish the attractiveness of the iPhone to end users.
Indeed, you could imagine a counterfactual where another judge in another universe decided that Apple should actually lock down the App Store even further, and charge an even higher commission: I actually think that this would have no meaningful difference on the perceived number of apps or on overall iPhone sales. Sure, developers would suffer, and some number of apps would prove to be unviable or, particularly in the case of apps that depend on advertising for downloads, less successful given their decreased ROAS, but the market is so large and liquid that the overall user experience and perceived value of apps would be largely the same.
This stark reality does, perhaps surprisingly, give me some amount of sympathy for Apple’s App Store intransigence. The fact of the matter is that everyone demanding more leniency in the App Store, whether that be in terms of commission rates or steering provisions or anything else, are appealing to nothing more than Apple’s potential generosity. The company’s self interest — and fiduciary duty to shareholders — has been 100% on the side of keeping the App Store locked down and commissions high.
Products → Platforms
That, by extension, is what bums me out about this entire affair. I would prefer to live in the tech world I and so many others mythologize, where platforms enable developers to make new apps, with those apps driving value to the underlying platform and making it even more attractive and profitable. That is what happened with the PC, and the creation of applications like VisiCalc and Photoshop.
That was also a much smaller market than today. VisiCalc came out in 1979, when 40,000–50,000 computers were sold; Photoshop launched on the Mac in 1990, with an addressable market of around a million Macs. The Vision Pro, meanwhile, is a flop for having sold only 500,000 units in 2024, nowhere near enough to attract a killer app, even if Apple’s App Store policies were not a hindrance.
None other than Meta CEO Mark Zuckerberg seems to recognize this new reality; one of my longest running critiques of Zuckerberg has been his continual obsession with building a platform a la Bill Gates and Windows, but as he told me last week, that’s not necessarily the primary goal now, even for Quest devices:
If you continue to deliver on [value] long term, is it still okay if that long term doesn’t include a platform, if you’re just an app?
MZ: It depends on what you’re saying. I think early on, I really looked up to Microsoft and I think that that shaped my thinking that, “Okay, building a developer platform is really cool”.
It is cool.
MZ: Yeah, but it’s not really the kind of company fundamentally that we have been historically. At this point, I actually see the tension between being primarily a consumer company and primarily a developer company, so I’m less focused on that at this point.
Now, obviously we do have developer surfaces in terms of all the stuff in Reality Labs, our developer platforms. We need to empower developers to build the content to make the devices good. The Llama stuff, we obviously want to empower people to use that and get as much of the world on open source as possible because that has this virtuous flywheel of effects that make it so that the more developers that are using Llama, the more Nvidia optimizes for Llama, the more that makes all our stuff better and drives costs down, because people are just designing stuff to work well with our systems and making their efficiency improvements to that. So, that’s all good.
But I guess the thing that I really care about at this point is just building the best stuff and the way to do that, I think, is by doing more vertical integration. When I think about why do I want to build glasses in the future, it’s not primarily to have a developer platform, it’s because I think that this is going to be the hardware platform that delivers the best ability to create this feeling of presence and the ultimate sense of technology delivering a social connection and I think glasses are going to be the best form factor for delivering AI because with glasses, you can let your AI assistant see what you see and hear what you hear and talk in your ear throughout the day, you can whisper to it or whatever. It’s just hard to imagine a better form factor for something that you want to be a personal AI that kind of has all the context about your life.
The great irony of Zuckerberg’s evolution — which he has been resisting for over a decade — is that this actually makes it more likely he will get a platform in the end. It seems clear in retrospect that DOS/Windows was the exception, not the rule; platforms, at least when it comes to the consumer space, are symptoms of products that move the needle. The only way to be a platform company is to be a product company first, and acquire the users that incentivize developers.
Takings and the Public Interest
Notice, however, the implication of this reality: an honest accounting of modern platforms, including iOS, is not simply that Apple, or whoever the platform provider is, owns intellectual property for which they have a right to be compensated; as I noted on Friday Apple has a viable appeal predicated on arguing Judge Gonzalez Rogers is “taking” their IP without compensation. What they actually own that is of the most value is user demand itself; to put it another way, Apple could charge a high commission and have stringent rules because developers wanted to be on their platform regardless. Demand draws supply, no matter the barriers.
This, in the end, is the oddity of this case, and the true “takings” violation: what is actually being taken from Apple is simply money. I don’t think anything is going to change about iPhone sales or app maker motivation; the former will simply be less profitable, and the latter more so. Small wonder Apple has fought to keep its position so strenuously!
This also leaves me more conflicted about Judge Gonzalez Rogers’ decision than I expected: I don’t like depriving Apple of their earned rewards, or diminishing the incentive to pursue this most difficult of goals — building a viable platform — in any way. Yes, Apple has made tons of money on the App Store, but the iPhone and associated ecosystem is of tremendous value.
At the same time, everything is a trade-off, and the fact that products that produce demand are the key to creating platforms significantly increases the public interest in regulating platform policies. I don’t think it is to society’s benefit to effectively delegate all innovation to platform providers, given how few there inevitably are; what needs incentivizing is experimentation on top of platforms, and that means recognizing that modern platform providers are in fact incentivized to tax that out of existence.
In fact you could, from a business perspective, make the case that Microsoft’s biggest mistake with Windows, at least from a shareholder perspective, was actually not harvesting as much value as they should have: two-sided network effects are so powerful that, once established, you can skim off as much money as you want with no ill effects. Apple certainly showed that was the case with the iPhone, and Google followed them; Meta has similar policies for Quest.
To that end, Congress should be prepared to act if Judge Gonzalez Rogers’ order is overturned on appeal; in fact, they should act anyways. My proposed law is clear and succinct:
A platform is a product with an API that runs 3rd-party applications.
A platform has 25 million+ U.S. users.
3rd-party applications should have the right, but not the compulsion, to (1) conduct commerce as they choose and (2) publish speech as they choose.
That’s it! If you want the benefit of 3rd-party applications (which are real — there’s an app for that!) then you have to offer fundamental economic and political freedom. This is in the American interest for the exact same reason that this wouldn’t kill the incentive to build the sort of product that leads to a platforms in the first place: platforms are so powerful that everyone in tech has, for decades, been both obsessed with them even as they underrated them.
You can subscribe to Netflix using the App Store by downloading one of Netflix’s games. ↩
Apple is not doomed, although things were feeling pretty shaky a couple of weeks ago, when the so-called “Liberation Day” tariffs were poised to make the company’s manufacturing model massively more expensive; the Trump administration granted Apple a temporary reprieve, and, for the next couple of months, things are business as usual.
Of course that’s not Apple’s only problem: a month ago the company had to admit that it couldn’t deliver on the AI promises it made at last year’s WWDC, leading John Gruber to declare that Something Is Rotten in the State of Cupertino. Still, just because Apple can’t ship a Siri that works is not necessarily cause for short-term concern: one of the Siri features Apple did ship was its ChatGPT integration, and you can run all of the best models as apps on your iPhone.
So no, Apple is not doomed, at least not for now. There is, however, real cause for concern: just as tech success is built years in advance, so is failure, and there are three historical examples of once-great companies losing the future that Apple and its board ought to consider carefully.
Microsoft and the Internet
I bet you think you already know the point I’m going to make here: Microsoft and the Internet is like Apple and AI. And you would be right! What may surprise you, however, is that I think this is actually good news for Apple, at least in part.
The starting point for the Internet is considered to be either 1991, when Tim Berners-Lee created the World Wide Web, or 1993 when Mosaic released the first consumer-accessible browser. In other words, Bill Gates’ famous memo about The Internet Tidal Wave was either two or four years late. This is from his opening:
Developments on the Internet over the next several years will set the course of our industry for a long time to come. Perhaps you have already seen memos from me or others here about the importance of the Internet. I have gone through several stages of increasing my views of its importance. Now I assign the Internet the highest level of importance. In this memo I want to make clear that our focus on the Internet is crucial to every part of our business. The Internet is the most important single development to come along since the IBM PC was introduced in 1981. It is even more important than the arrival of the graphical user interface (GUI). The PC analogy is apt for many reasons. The PC wasn’t perfect. Aspects of the PC were arbitrary or even poor. However a phenomena grew up around the IBM PC that made it a key element of everything that would happen for the next 15 years. Companies that tried to fight the PC standard often had good reasons for doing so but they failed because the phenomena overcame any weaknesses that resisters identified.
It’s unfair to call this memo “late”: it’s actually quite prescient, and Microsoft pivoted hard into the Internet — so hard that just a few years later they faced a DOJ lawsuit that was primarily centered around Internet Explorer. In fact, you could make a counterintuitive argument that Microsoft actually suffered from Gates’ prescience; this was what he wrote about Netscape:
A new competitor “born” on the Internet is Netscape. Their browser is dominant, with 70% usage share, allowing them to determine which network extensions will catch on. They are pursuing a multi-platform strategy where they move the key API into the client to commoditize the underlying operating system. They have attracted a number of public network operators to use their platform to offer information and directory services. We have to match and beat their offerings including working with MCI, newspapers, and other who are considering their products.
Microsoft beat Netscape, but to what end? The client was in fact commoditized — the Internet Explorer team actually introduced the API that made web apps possible — but that was OK for business because everyone used Windows already.
What actually mattered was openness, in two regards: first, because the web was open, Microsoft ultimately could not contain it to just its platform. Second, because Windows was open, it didn’t matter: Netscape, to take the most pertinent example, was a Windows app; so was Firefox, which dethroned Internet Explorer after Microsoft lost interest, and so is Chrome, which dominates the web today.
That’s not to say that the Internet didn’t matter to Microsoft’s long-term prospects, because it was a bridge to the paradigm that Microsoft actually fumbled, which was mobile. Last fall I wrote The Gen AI Bridge to the Future, where I made the argument that paradigm shifts in hardware were enabled by first building “bridges” at the application layer. Here is the section on Windows and the Internet:
PCs underwent their own transformation over their two decades of dominance, first in terms of speed and then in form factor, with the rise of laptops. The key innovation at the application layer, however, was the Internet:
The Internet differed from traditional applications by virtue of being available on every PC, facilitating communication between PCs, and by being agnostic to the actual device it was accessed on. This, in turn, provided the bridge to the next device paradigm, the smartphone, with its touch interface:
I’ve long noted that Microsoft did not miss mobile; their error was in trying to extend the PC paradigm to mobile. This not only led to a focus on the wrong interface (WIMP via stylus and built-in keyboard), but also an assumption that the application layer, which Windows dominated, would be a key differentiator.
Apple, famously, figured out the right interface for the smartphone, and built an entirely new operating system around touch. Yes, iOS is based on macOS at a low level, but it was a completely new operating system in a way that Windows Mobile was not; at the same time, because iOS was based on macOS, it was far more capable than smartphone-only alternatives like BlackBerry OS or PalmOS. The key aspect of this capability was that the iPhone could access the real Internet… that was the key factor in reinventing the phone, because it was the bridge that linked a device in your pocket to the world of computing writ large.
To reiterate Microsoft’s failure, the company attempted to win in mobile by extending the Windows interface and applications to smartphones; what the company should have done is “pursu[e] a multi-platform strategy where they move the key API into the client to commoditize the underlying operating system.” In other words, Microsoft should have embraced and leveraged the Netscape threat, instead of trying to neutralize it.
Apple and the iPhone is analogous to Microsoft and Windows, for better and for worse: the better part is that there are many more smartphones sold than PCs, which means that Apple, even though it controls less than half the market, has more iOS devices than there are Windows devices. That’s the “for worse” part, however: Apple exerts more control on iOS than Microsoft ever did on Windows, but also doesn’t have a monopoly like Microsoft did.
The most obvious consequence of smartphones being a duopoly is that Apple can’t unilaterally control the entire industry’s layer like Microsoft wanted to. However, you can look at this in a different way: Microsoft couldn’t have dared to exert Apple-like control of Windows because it was a monopoly; the Windows API was, as I noted above, an open one, and that meant that the Internet largely happened on Windows PCs.
Consider this in the context of AI: the iPhone does have AI apps from everyone, including ChatGPT, Claude, Gemini, DeepSeek, etc. The system-wide assistant interface, however, is not open: you’re stuck with Siri. Imagine how much more attractive the iPhone would be as an AI device if it were a truly open platform: the fact that Siri stinks wouldn’t matter, because everyone would be running someone else’s model.
Where this might matter more is the next device paradigm: the point of The Gen AI Bridge to the Future is in the title:
We already established above that the next paradigm is wearables. Wearables today, however, are very much in the pre-iPhone era. On one hand you have standalone platforms like Oculus, with its own operating system, app store, etc.; the best analogy is a video game console, which is technically a computer, but is not commonly thought of as such given its singular purpose. On the other hand, you have devices like smart watches, AirPods, and smart glasses, which are extensions of the phone; the analogy here is the iPod, which provided great functionality but was not a general computing device.
Now Apple might dispute this characterization in terms of the Vision Pro specifically, which not only has a PC-class M2 chip, along with its own visionOS operating system and apps, but can also run iPad apps. In truth, though, this makes the Vision Pro akin to Microsoft Mobile: yes, it is a capable device, but it is stuck in the wrong paradigm, i.e. the previous one that Apple dominated. Or, to put it another way, I don’t view “apps” as the bridge between mobile and wearables; apps are just the way we access the Internet on mobile, and the Internet was the old bridge, not the new one.
The new bridge is a user interface that gives you exactly what you need when you need it, and disappears otherwise; it is based on AI, not apps. The danger for Apple is that trying to keep AI in a box in its current paradigm will one day be seen like Microsoft trying to keep the Internet locked to its devices: fruitless to start, and fatal in the end.
Intel and the Foundry Model
Intel was the other company that dominated the PC era: while AMD existed, they were more of an annoyance than an actual threat (thanks in part to Intel’s own anticompetitive behavior). And, like Microsoft, Intel also missed mobile, for somewhat similar reasons: they were over-indexed on the lessons of the PC.
Back in the 1980s and 1990s, when PCs were appearing on every desk and in every home, the big limitation was performance; Intel, accordingly, was focused on exactly that: every generation of Intel chips was massively faster than the previous one, and the company delivered so regularly that developers learned to build for the future, and not waste time optimizing for the soon-to-be-obsolete present.
Mobile, however, meant battery power, and Intel just wasn’t that concerned about efficiency; while the popular myth is that Intel turned Apple down when it came to building chips for the iPhone, Tony Fadell told me in a Stratechery Interview that they were never under consideration:
The new dimension that always came in with embedded computing was always the power element, because on battery-operated devices, you have to rethink how you do your interrupt structures, how you do your networking, how you do your memory. You have to think about so many other parameters when you think about power and doing enough processing effectively, while having long battery life. So everything for me was about long, long battery life…when you take that microscopic view of what you’re building, you look at the world very differently.
For me, when it came to Intel at the time, back in the mid-2000s, they were always about, “Well, we’ll just repackage what we have on the desktop for the laptop and then we’ll repackage that again for embedding.” It reminded me of Windows saying, “I’m going to do Windows and then I’m going to do Windows Mobile and I’m going to do Windows embedded.” It was using those same cores and kernels and trying to slim them down…”We’re just going to have Moore’s Law take over” and so in a way that locks you into a path and that’s why Intel, not under the Pat days but previous to the Pat days, was all driven by manufacturing capability and legal. It wasn’t driven by architectural decisions.
Missing mobile was a big problem for Intel’s integrated device manufacturing model: the company, in the long run, would not have the volume and the associated financial support of mobile customers to keep up with TSMC. Today the company is struggling to turn itself into a foundry — a company that manufactures chips for external customers — and would like nothing more than to receive a contract from the likes of Apple, not for an Intel chip, but for an ARM-based one.
What is notable about this example, however, is how long it took to play out. One of my first Articles on Stratechery was 2013’s The Intel Opportunity, where I urged the company to get into the foundry business, a full six years after the iPhone came out; I thought I was late. In fact, Intel’s stock nearly reached its dot-com era highs in 2020, after steady growth in the seven years following that Article:
The reason for that growth was, paradoxically enough, mobile: the rise of smartphones was mirrored by the rise of cloud computing, for which Intel made the processors. Better yet, those Xeon processors were much more expensive than PC processors (much less mobile ones), which meant margins kept growing; investors didn’t seem to care that Intel’s decline — so apparent today — was already locked in.
While Microsoft and the Internet is more directly analogous to Apple and AI, it’s the collective blindness of Intel shareholders and management to the company’s long-term risks that offers a lesson for the iPhone maker. To summarize the Intel timeline:
Intel missed mobile because it was focused on the wrong thing (performance over efficiency).
Intel failed to leverage its greatest strength (manufacturing) into an alternative position in mobile (being a foundry).
Intel’s manufacturing fell behind the industry’s collective champion (TSMC), which raised challenges to Intel’s core business (AMD server chips are now better than Intel’s).
Now, a decade-and-a-half after that first mistake, Intel is on the ropes, despite all of the money it made and stock market increases it enjoyed in the meantime.
If a similar story unfolds for Apple, it might look like this:
Apple misses AI because it’s focused on the wrong thing (privacy).
Apple fails to leverage its greatest strength (the iPhone platform) into an alternative position in AI (being the platform for the best model makers).
Apple’s platform falls behind the industry’s collective champion (Android or perhaps TBD), which raises challenges to Apple’s core business (AI is so important that the iPhone has a worse user experience).
The questions about Apple’s privacy focus being a hindrance in AI are longstanding ones; I raised them in this 2015 Update when I noted that the company’s increasingly strident stance on data collection ran the risk of diminishing product quality as machine learning rose in importance.
In fact, those fears turned out to be overblown for a good long while; many would argue that Apple’s stance (strategy credit or not) was a big selling point. I think it’s fair to wonder, however, if those concerns were not wrong but simply early:
An Apple completely unconcerned with privacy would have access to a vast trove of exclusive user data on which to train models.
An Apple that refused to use user data for training could nonetheless deliver a superior experience by building out its AI as a fully scaled cloud service, instead of the current attempt to use on-device processing and a custom-built private cloud compute infrastructure that, by necessity, has to rely on less capable models and worse performance.
An Apple that embraced third party model providers could, as noted above, open up its operating systems so that users could replace Siri with the model of their choice.
Apple’s absolutist and paternalistic approach to privacy have taken all of these options off the table, leaving the company to provide platform-level AI functionality on its own with a hand tied behind its back, and to date the company has not been able to deliver; given how different AI is than building hardware or operating systems, it’s fair to wonder if they ever will.
And, critically, this won’t matter for a long time: Apple’s AI failures will not impact iPhone sales for years, and most AI use cases will happen in apps that run on the iPhone. What won’t happen, however, is the development of the sort of platform capabilities that will build that bridge to the future.
This, in the end, was Intel’s ultimate failing: today there is massive demand for foundry capacity, but not for mobile; what the world wants is more AI chips, particularly from a company (Nvidia) which has regularly been willing to dual source its supply. Intel, though, has yet to meet the call; the cost of the company not opening itself up after its mobile miss is that it wasn’t prepared for the next opportunity that came along.
Apple and China
This last analogy is, I admit, the shakiest, but perhaps the most important: it’s Apple itself. From the New York Times:
In 1983, Mr. Jobs oversaw the construction of a state-of-the-art plant where the new Macintosh computer would be built. Reporters who toured it early on were told that the plant, located just across San Francisco Bay from Apple’s headquarters, was so advanced that factory labor would account for 2 percent of the cost of making a Macintosh. Ultimately, the Macintosh factory closed in 1992, in part because it never realized the production volume that Mr. Jobs had envisioned — such sales numbers for the Mac would only come later…
That failure taught Mr. Jobs the lesson. He returned to Apple in 1997, and the next year, he hired Tim Cook as Apple’s senior vice president for worldwide operations. Mr. Cook had mastered the art of global manufacturing supply chains, first in IBM’s personal computer business and then at Compaq Computer.
It was admirable that Jobs wanted to build in America, but realistically the company needed to follow the rest of the tech industry to Asia if it wanted to survive, much less thrive, and Cook, just as much as Jobs, both saved the company and set it on the course for astronomical growth.
The challenge today is that that growth has been mirrored by China itself, and the current administration is determined to decouple the U.S. from China; that potentially increases Apple’s most existential threat, which is a war over Taiwan. This is a very different problem than what has long concerned Cook; from a 2008 profile in Fortune:
Almost from the time he showed up at Apple, Cook knew he had to pull the company out of manufacturing. He closed factories and warehouses around the world and instead established relationships with contract manufacturers. As a result, Apple’s inventory, measured by the amount of time it sat on the company’s balance sheet, quickly fell from months to days. Inventory, Cook has said, is “fundamentally evil,” and he has been known to observe that it declines in value by 1% to 2% a week in normal times, faster in tough times like the present. “You kind of want to manage it like you’re in the dairy business,” he has said. “If it gets past its freshness date, you have a problem.” This logistical discipline has given Apple inventory management comparable with Dell’s, then as now the gold standard for computer-manufacturing efficiency.
There are things worse than dairy going bad: it’s cows being blown up. Evil? Absolutely. Possible? Much more so today than at any other point in Cook’s tenure.
This, then, is the analogy: the Apple that Cook arrived at in 1998 was at existential risk from its supply chain; so is Apple today. Everything else is different, including the likelihood of disaster; Apple’s China risk may be elevated, whereas Apple’s bankruptcy in the 1990’s seemed a matter of when, not if:
At the same time, that also means that Apple has cash flow, and power; what is necessary now is not making obvious choices out of necessity, but making uncertain ones out of prudence. Cook built the Apple machine in China; the challenge now will be in dismantling it.
The Cook Question
Cook is the common variable across all of these analogies:
Cook has led the company as it has continually closed down iOS, controlling developers through the stick of market size instead of the carrot of platform opportunity.
Cook has similarly been at the forefront of Apple’s absolutist approach to privacy, which has only increased in intensity and impact, not just on 3rd parties but also on Apple itself.
Cook, as I just documented, built Apple’s dependency on China, and has adroitly managed the politics of that reality, both with China and the U.S.
All of these decisions — even the ones I have most consistently disagreed with — were defensible and, in some cases, essential to Apple’s success; Cook has been a very effective CEO for Apple and its shareholders. And, should he stay on for several more years, the company would probably seem fine (assuming nothing existential happens with China and Taiwan), particularly in terms of the stock price.
Tech fortunes, however, are cast years in advance; Apple is not doomed, but it is, for the first time in a long time, fair to wonder about the long-term: the questions I have about the company are not about 2025, but 2035, and the decisions that will answer those questions will be made now. I certainly have my point of view:
Apple should execute an AI Platform Pivot, enabling developers to build with AI instead of trying to do everything itself; more broadly, it should increase the opportunities for developers economically and technically.
Apple should not abandon its privacy brand, but rather accept the reality that all of computing is ultimately about trust: the device will always have root. To that end, users do trust Apple, not because Apple is so strident about user data that they make their products worse, but because the company’s business model is aligned with users, with a multi-decade track record of doing right by them; in this case, doing right by users means doing what is necessary to have an actually useful AI offering.
Whereas I once thought it was reasonable for Apple to maintain its position in China — the costs of hedging would be so large that it would be better to take the minuscule risk of war, which Apple itself minimized through its position in China — that position no longer seems feasible; at a minimum Apple needs to rapidly accelerate its diversification efforts. This doesn’t just mean building up final assembly in places like India and Brazil, but also reversing its long-running attempts to undercut non-Chinese suppliers with Chinese alternatives.
All of these run counter to the decisions Cook has made over the last three decades, but again, it’s not that Cook was wrong at the time he made them; rather, times change, and Apple needs to change before the time comes where the necessity for change is obvious, because that means the right time for that change has already passed.
I have a few informal guidelines that govern my writing on Stratechery, including “Don’t post more than one front-page Article a week”, “Don’t talk about my writing process”, and “Don’t start Articles with ‘I’”; it’s an extraordinary week, though, so I’m breaking a few rules.
There are three old Stratechery Articles that, after reflection, missed the mark in different ways.
The most proximate cause for my rule-breaking was Monday’s Trade, Tariffs, and Tech; I stand by everything I wrote, but it was incomplete, lacking an overall framework and satisfying conclusion. That’s not surprising given the current uncertainty, but that means I should have waited to publish a front-page Article until I had more clarity (I do much more musing in my Updates, which that Article is now categorized as). Now I have to break my rule and write another Article.
The second Article to revisit is November’s A Chance to Build. This Article was in fact deeply pessimistic about President Trump’s promised trade regime, particularly in terms of what it meant for tech; the title and conclusion, however, tried to find some positives. Clearly that was a mistake; that Article was predictive of what was happening, but I obscured the prediction.
The third Article to revisit is January 2021’s Internet 3.0 and the Beginning of (Tech) History. This Article was right about tech exiting an economically-defined era — the Aggregation era — and entering a new politically-defined era. It was, however, four years too early, and misdiagnosed the reason for the transition. The driver is not foreign countries closing their doors to America; it’s America closing its door to the world.
The proximate cause of all of this reflection is of course Trump’s disastrous “liberation day” tariffs. The secondary cause is what I wrote about Monday: the U.S. has a genuine problem on its hands thanks to its inability to make things pertinent to modern warfare and high tech. The root cause, however, is very much in Stratechery’s wheelhouse, and worthy of another Article: it’s disruption.
“Disruption” describes a process whereby a smaller company with fewer resources is able to successfully challenge established incumbent businesses. Specifically, as incumbents focus on improving their products and services for their most demanding (and usually most profitable) customers, they exceed the needs of some segments and ignore the needs of others. Entrants that prove disruptive begin by successfully targeting those overlooked segments, gaining a foothold by delivering more-suitable functionality—frequently at a lower price. Incumbents, chasing higher profitability in more-demanding segments, tend not to respond vigorously. Entrants then move upmarket, delivering the performance that incumbents’ mainstream customers require, while preserving the advantages that drove their early success. When mainstream customers start adopting the entrants’ offerings in volume, disruption has occurred.
This is almost a perfect summary of what has happened in manufacturing, and, as I noted in that November article, it started with chips:
That history starts in 1956, when William Shockley founded the Shockley Semiconductor Laboratory to commercialize the transistor that he had helped invent at Bell Labs; he chose Mountain View to be close to his ailing mother. A year later the so-called “Traitorous Eight”, led by Robert Noyce, left and founded Fairchild Semiconductor down the road. Six years after that Fairchild Semiconductor opened a facility in Hong Kong to assemble and test semiconductors. Assembly required manually attaching wires to a semiconductor chip, a labor-intensive and monotonous task that was difficult to do economically with American wages, which ran about $2.50/hour; Hong Kong wages were a tenth of that. Four years later Texas Instruments opened a facility in Taiwan, where wages were $0.19/hour; two years after that Fairchild Semiconductor opened another facility in Singapore, where wages were $0.11/hour.
In other words, you can make the case that the classic story of Silicon Valley isn’t completely honest. Chips did have marginal costs, but that marginal cost was, within single digit years of the founding of Silicon Valley, exported to Asia.
Notice what did still happen in the United States, at least back then: actual chip fabrication. That was where innovation happened, and where margins were captured, so of course U.S. chip companies kept that for themselves. It was the tedious and labor-intensive assembly and testing that was available to poor Asian economies led by authoritarian governments eager to provide some sort of alternative to communism.
One important point about new market disruption — which Asian manufacturing was — is that it is downstream of a technological change that fundamentally changes cost structures. In the case of the Asian manufacturing market, there were actually three; from 2016’s The Brexit Possibility:
In the years leading up to the 1970s, three technological advances completely transformed the meaning of globalization:
In 1963 Boeing produced the 707-320B, the first jet airliner capable of non-stop service from the continental United States to Asia; in 1970 the 747 made this routine.
In 1964 the first transpacific telephone cable between the United States and Japan was completed; over the next several years it would be extended throughout Asia.
In 1968 ISO 668 standardized shipping containers, dramatically increasing the efficiency with which goods could be shipped over the ocean in particular.
These three factors in combination, for the first time, enabled a new kind of trade. Instead of manufacturing products in the United States (or Europe or Japan or anywhere else) and trading them to other countries, multinational corporations could invert themselves: design products in their home markets, then communicate those designs to factories in other countries, and ship finished products back to their domestic market. And, thanks to the dramatically lower wages in Asia (supercharged by China’s opening in 1978), it was immensely profitable to do just that.
Christensen, somewhat confusingly, actually has two theories of disruption; the other one is called “low-end disruption”, but it is also pertinent to this story. From The Innovator’s Solution:
The pressure of competing along this new trajectory of improvement [(speed, convenience, and customization)] forces a gradual evolution in product architecture, as depicted in Figure 5-1 — away from the interdependent, proprietary architectures that had the advantage in the not-good-enough era toward modular designs in the era of performance surplus. Modular architectures help companies to compete on the dimensions that matter in the lower-right portions of the disruption diagram. Companies can introduce new products faster because they can upgrade individual subsystems without having to redesign everything. Although standard interfaces invariably force compromise in system performance, firms have the slack to trade away some performance with these customers because functionality is more than good enough.
Modularity has a profound impact on industry structure because it enables independent, nonintegrated organizations to sell, buy, and assemble components and subsystems. Whereas in the interdependent world you had to make all of the key elements of the system in order to make any of them, in a modular world you can prosper by outsourcing or by supplying just one element. Ultimately, the specifications for modular interfaces will coalesce as industry standards. When that happens, companies can mix and match components from best-of-breed suppliers in order to respond conveniently to the specific needs of individual customers. As depicted in Figure 5-1, these nonintegrated competitors disrupt the integrated leader.
This is exactly what happened to categories like PCs: everything became modular, commoditized, and low margin — and thus followed chip test and assembly to Asia. One aspect that was under-discussed in Christensen’s theory, however, was scale, which mattered more than the customization point. It was less important that a customer be able to use any chip they wanted than it was that a lot of customers wanted to use the same chip. Moreover, this scale point applied up-and-down the stack, to both components and assemblers.
Note also the importance of scale to the new market disruption above: while outsourcing got easier thanks to technology, it’s difficult to be easier than working locally; the best way to overcome those coordination costs is to operate at scale. This helps explain why manufacturing in Asia is fundamentally different than the manufacturing we remember in the United States decades ago: instead of firms with product-specific factories, China has flexible factories that accommodate all kinds of orders, delivering on that vector of speed, convenience, and customization that Christensen talked about.
This scale has, as I noted last November, been particularly valuable for tech companies; software scales to the world, and Asian factories, particularly Chinese ones, scale with it, providing the hardware complements to American software. That is why every single tech company — even software ones — is damaged by these tariffs; more expensive complements means lower usage overall.
The other scale point that is particularly pertinent to technology is chips. Every decrease in node size comes at increasingly astronomical costs; the best way to afford those costs is to have one entity making chips for everyone, and that has turned out to be TSMC. Indeed, one way to understand Intel’s struggles is that it was actually one of the last massive integrated manufacturers: Intel made chips almost entirely for itself. However, once the company missed mobile, it had no choice but to switch to a foundry model; the company is trying now, but really should have started fifteen years ago. Now the company is stuck, and I think they will need government help.
iPhone Jobs
There is one other very important takeaway from disruption: companies that go up-market find it impossible to go back down, and I think this too applies to countries. Start with the theory: Christensen had a chapter in The Innovator’s Dilemma entitled “What Goes Up, Can’t Go Down”:
Three factors — the promise of upmarket margins, the simultaneous upmarket movement of many of a company’s customers, and the difficulty of cutting costs to move downmarket profitably — together create powerful barriers to downward mobility. In the internal debates about resource allocation for new product development, therefore, proposals to pursue disruptive technologies generally lose out to proposals to move upmarket. In fact, cultivating a systematic approach to weeding out new product development initiatives that would likely lower profits is one of the most important achievements of any well-managed company.
Now consider this in the context of the United States: every single job in this country, even at the obsolete federal minimum wage of $7.25/hour, makes much more money than an iPhone factory line worker. And, critically, we have basically full employment; that is what makes this statement from White House Press Secretary Karoline Leavitt ridiculous; from 9to5Mac:
In response to a question from Maggie Haberman of The New York Times about the types of jobs Trump hopes to create in the U.S. with these tariffs, Leavitt said:
“The president wants to increase manufacturing jobs here in the United States of America, but he’s also looking at advanced technologies. He’s also looking at AI and emerging fields that are growing around the world that the United States needs to be a leader in as well. There’s an array of diverse jobs. More traditional manufacturing jobs, and also jobs in advanced technologies. The president is looking at all of those. He wants them to come back home.”
Haberman followed up with a question about iPhone manufacturing specifically, asking whether Trump thinks this is “the kind of technology” that could move to the United States. Leavitt responded:
“[Trump] believes we have the labor, we have the workforce, we have the resources to do it. As you know, Apple has invested $500 billion here in the United States. So, if Apple didn’t think the United States could do it, they probably wouldn’t have put up that big chunk of change.”
So could Apple pay more to get U.S. workers? I suppose — leaving aside the questions of skills and whatnot — but there is also the question of desirability; the iPhone assembly work that is not automated is highly drudgerous, sitting in a factory for hours a day delicately assembling the same components over and over again. It’s a good job if the alternative is working in the fields or in a much more dangerous and uncomfortable factory, but it’s much worse than basically any sort of job that is available in the U.S. market.
At the same time, it is important to note that this drudgerous final assembly work is a center of gravity for the components that actually need to be assembled, and these parts are all of significantly higher value, and far more likely to be produced through automation. As I noted yesterday, Apple has probably done more than any other company to move China up the curve in terms of the ability to manufacture components, often to the detriment of suppliers in the U.S., Taiwan, South Korea, Japan, etc.; from Apple’s perspective spending time and money to bring Chinese component suppliers online provides competition for its most important suppliers, giving them greater negotiating leverage. From the U.S.’s perspective this means that a host of technologies and capabilities downstream from the smartphone — which is to say nearly all electronics, including those with significant military applicability like drones — are being developed in China.
Beyond Disruption
Fortunately, while true disruption is often the ultimate death knell for an individual company with a specific value proposition, I don’t think it is a law of nature. Disruption is about supply, but success on the Internet, to take one example familiar to Stratechery readers, is about demand — and controlling demand is more important than controlling supply. I expanded on this in a 2015 Article called Beyond Disruption:
The Internet has completely transformed business by making both distribution and transaction costs effectively free. In turn, this has completely changed the calculus when it comes to adding new customers: specifically, it is now possible to build businesses where every incremental customer has both zero marginal costs and zero opportunity costs. This has profound implications: instead of some companies serving the high end of a market with a superior experience while others serve the low-end with a “good-enough” offering, one company can serve everyone. And, given the choice between a superior experience and one that is “good-enough,” of course the superior experience will win.
To be sure, it takes time to scale such a company, but given the end game of owning the entire market, the rational approach is not to start on the low-end, but rather the exact opposite. After all, while marginal costs may be zero, providing a superior experience in the age of the Internet entails significant upfront (fixed) costs, and while those fixed costs are minimized on a per-customer basis at scale, they can have a significant impact with a small customer base. Therefore, it makes sense to start at the high-end with customers who have a greater willingness-to-pay, and from there scale downwards, decreasing your price along with the decrease in your per-customer cost base (because of scale) as you go (and again, without accruing material marginal costs).
This is exactly what Uber has done: the company spent its early years building its core technology and delivering a high-end experience with significantly higher prices than incumbent taxi companies. Eventually, though, the exact same technology was deployed to deliver a lower-priced experience to a significantly broader customer base; said customer base was brought on board at zero marginal cost.
I want to be careful not to draw too many lessons from Aggregation Theory in an Article about manufacturing, given there are by definition marginal costs involved in physical goods. However, I would note two things:
First, marginal manufacturing costs are, for many goods, going down over time, thanks to automation; indeed, this is why the U.S. still has a significant amount of manufacturing output even if an ever-decreasing number of people are employed in the manufacturing sector.
Second, the idea that demand matters most does still hold. The takeaway from that Article isn’t that Uber is a model for the rebirth of American manufacturing; rather it’s that you can leverage demand to fundamentally reshape supply.
It’s not as if the Trump administration doesn’t know this: the entire premise of these tariffs is that everyone wants access to the U.S. market, and rightly so given the outsized buying power driven both by our wealth and by the capacity for borrowing afforded us by the dollar being the reserve currency. It’s also true that China has an excess of supply; given that supply is usually built with debt that means the country needs cash flow, and even if factories are paid off, the country needs the employment opportunties. China’s hand is not as strong as many of Trump’s strongest critics believe.
The problem with these tariffs is that their scale and indiscriminate nature will have the effect of destroying demand and destroying the capability to develop alternative supply. I suppose if the only goal is to hurt China then shooting yourself in the foot, such that you no longer need to buy shoes for stumps, is a strategy you could choose, but that does nothing to help with what should be the primary motivation: shoring up the U.S. national security base.
Those national security concerns are real. The final stage of disruption is when the entity that started on the bottom is uniquely equipped to deliver what is necessary for a new paradigm, and that is exactly what happened with electronics generally and drones specifically. Moreover, this capability is only going to grow more important with the rise of AI, which will be substantiated in the physical world through robotics. And, of course, robots will be the key to building other robots; if the U.S. wants to be competitive in the future, and not be dependent on China, it really does need to make changes — just not these ones.
A Better Plan
The key distinguishing feature of a better plan is that it doesn’t seek to own supply, but rather control it in a way the U.S. does not today.
First, blanket tariffs are a mistake. I understand the motivation: a big reason why Chinese imports to the U.S. have actually shrunk over the last few years is because a lot of final assembly moved to countries like Vietnam, Thailand, Mexico, etc. Blanket tariffs stop this from happening, at least in theory.
The problem, however, is that those final assembly jobs are the least desirable jobs in the value chain, at least for the American worker; assuming the Trump administration doesn’t want to import millions of workers — that seems rather counter to the foundation of his candidacy! — the United States needs to find alternative trustworthy countries for final assembly. This can be accomplished through selective tariffs (which is exactly what happened in the first Trump administration).
Secondly, using trade flows to measure the health of the economic relationship with these countries — any country, really, but particularly final assembly countries — is legitimately stupid. Go back to the iPhone: the value-add of final assembly is in the single digit dollar range; the value-add of Apple’s software, marketing, distribution, etc. is in the hundreds of dollars. Simply looking at trade flows — where an imported iPhone is calculated as a trade deficit of several hundred dollars — completely obscures this reality. Moreover, the criteria for a final assembly country is that they have low wages, which by definition can’t pay for an equivalent amount of U.S. goods to said iPhone.
At the same time, the overall value of final assembly does exceed its economic value, for the reasons noted above: final assembly is gravity for higher value components, and it’s those components that are the biggest national security problem. This is where component tariffs might be a useful tool: the U.S. could use a scalpel instead of a sledgehammer to incentivize buying components from trusted allies, or from the U.S. itself, or to build new capacity in trusted locations. This does, admittedly, start to sound a lot like central planning, but that is why the gravity argument is an important one: simply moving final assembly somewhere other than China is a win — but not if there are blanket tariffs, at which point you might as well leave the supply chain where it is.
Third, the most important components for executing a fundamental shift in trade are those that go into building actual factories, or equipment for those factories. In the vast sea of stupidity that are these tariffs this is perhaps the stupidest detail of all: the U.S. is tariffing raw materials and components for factory equipment, like CNC machines. Consider this announcement from Haas:
Breaking – Haas reduces production of CNC machines, eliminates overtime, halts hiring, citing "dramatic decrease in demand" and concerns about reduced tariffs on Asian machines. Asks the administration for tariff exemptions on imported machine components and raw materials. pic.twitter.com/XOAnDWQhkx
You can certainly make the case that things like castings and other machine components are of sufficient importance to the U.S. that they ought to be manufactured here, but you have to ramp up to that. What is much more problematic is that raw materials and components are now much cheaper for Haas’ foreign competitors; even if those competitors face tariffs in the United States, their cost of goods sold will be meaningfully lower than Haas, completely defeating the goal of encouraging the purchase of U.S. machine tools.
I get the allure of blanket tariffs; politics is often the art of the possible, and the perfect is the enemy of the good. The problem is this approach simply isn’t good: it’s actively detrimental to what should be the U.S.’s goals. It’s also ignoring the power of demand: China would supply factories in the U.S., even if the point of those factories was to displace China, because supply needs to sell. This is how you move past disruption: you not only exert control on alternatives to China, you exert control on China itself.
Fourth, there remains the problem of chips. Trump just declared economic war on China, which definitionally increases the possibility of kinetic war. A kinetic war, however, will mean the destruction of TSMC, leaving the U.S. bereft of chips at the very moment that A.I. is poised to create tremendous opportunities for growth and automation. And, even if A.I. didn’t exist, it’s enough to note that modern life would grind to a halt without chips. That’s why this is the area that most needs direct intervention from the federal government, particularly in terms of incentivizing demand for both leading and trailing edge U.S. chips.
I do, as I noted on Monday, have more sympathy than many of Trump’s critics for the need to make fundamental changes to trade; that, however, doesn’t mean any change is ipso facto good: things could get a lot worse, and these “liberation day” tariffs will do exactly that.
The Melancholy of Internet 3.0
I started this essay being solipsistic, so let me conclude with some more navel-gazing: my prevailing emotion over the past week — one I didn’t fully come to grips with until interrogating why Monday’s Article failed to live up to my standards — is sadness over the end of an era in technology, and frustration-bordering-on-disillusionment over the demise of what I thought was a uniquely American spirit.
Internet 1.0 was about technology. This was the early web, when technology was made for technology’s sake. This was when we got standards like TCP/IP, DNS, HTTP, etc. This was obviously the best era, but one that was impossible to maintain once there was big money to be made on the Internet.
Internet 2.0 was about economics. This was the era of Aggregators — the era of Stratechery, in other words — when the Internet developed, for better or worse, in ways that made maximum economic sense. This was a massive boon for the U.S., which sits astride the world of technology; unfortunately none of the value that comes from that position is counted in the trade statistics, so the administration doesn’t seem to care.
Internet 3.0 is about politics. This is the era when countries make economically sub-optimal choices for reasons that can’t be measured in dollars and cents. In that Article I thought that Big Tech exercising its power against the President might be a spur for other countries to seek to wean themselves away from American companies; instead it is the U.S. that may be leaving other countries little choice but to retaliate against U.S. tech.
One can certainly make the case that the Internet 2.0 era wasn’t ideal, or even actively detrimental; it’s similar to the case that while free trade might have made everyone — especially the U.S. — richer, it wasn’t worth national security sacrifices that we are only now waking up to. For me, though, it was the era that has defined my professional life, and I’m sad to see it slipping away. Stratechery has always been non-political; it bums me out if we are moving to an era where politics are inescapable — they certainly are this week.
The second emotion — the frustration-bordering-on-disillusionment — is about the defeatist and backwards-looking way that the U.S. continues to approach China. These tariffs, particularly to the extent they are predicated on hurting China, are a great example: whether through malice or incompetence this particular tariff plan seems designed to inflict maximal pain, even though that means hurting the U.S. along the way. What is worse is that this is a bipartisan problem: Biden’s chip controls are similarly backwards looking, seeking to stay ahead by pulling up the ladder of U.S. technology, instead of trying to stay ahead through innovation.
There is, admittedly, a hint of that old school American can-do attitude embedded in these tariffs: the Trump administration seems to believe the U.S. can overcome all of the naysayers and skeptics through sheer force of will. That force of will, however, would be much better spent pursuing a vision of a new world order in 2050, not trying to return to 1950. That is possible to do, by the way, but only if you accept 1950’s living standards, which weren’t nearly as attractive as nostalgia-colored glasses paint them, and if we’re not careful, 1950’s technology as well. I think we can do better than that; I know we can do better than this.
While “March Madness” refers to the NCAA basketball tournaments, the maddest weekend of all is the first one, when fields of 641 are trimmed down to the Sweet 16; this means there are 16 games a day the first two days, and 8 games a day for the next two. Inevitably this means that multiple games are on at the same time, and Max has a solution for you; from The Streamable:
The 2025 NCAA Men’s Basketball Tournament starts today, and just in time, Warner Bros. Discovery has announced the addition of some very modern features for games that stream on its on-demand service Max. Fans can use Max to stream all March Madness games on TNT, TBS, and truTV, and that viewing experience is about to improve in a big way.
The new Max feature that fans will likely appreciate most while watching NCAA Men’s Basketball Tournament games is a multiview. This will allow fans to watch up to three games at once, ensuring they never miss a single bucket, block, or steal from the tournament.
Except that’s not correct; Warner Bros. Discovery shares the rights to the NCAA Men’s Basketball Tournament with CBS, and there were times over the weekend when there were games on CBS and a Warner Bros. Discovery property — sometimes four at once. That means that Max multiview watchers were in fact missing buckets, blocks, and steals, and likely from the highest profile games, which were more likely to be on the broadcast network.
Notice, however, that I specified Max multiview watchers; YouTube TV has offered multiview for the NCAA Tournament since last year. Critically, YouTube TV’s offering includes CBS, and, starting this upcoming weekend, will also let you watch the women’s tournament as well; from Sportico:
Generally, events from the same leagues are kept together. On Friday, for instance, men’s and women’s multiviews will be offered separately. If you truly want to watch all of March Madness live, it’ll be time to break out that second screen again. However, in part due to user demand, YouTube TV says mixed gender multiviews will be available starting with the Sweet 16.
The job of prioritizing selections has only gotten more complicated as interest in women’s hoops has boomed. Through the first two rounds in 2024, viewership of the women’s tourney was up 108% over the year prior. Though the “March Madness” brand is now used for both men’s and women’s competitions, separate media deals dictate their distribution. CBS and TNT Sports networks split the men’s games, including streaming on March Madness Live apps, while ESPN’s channels host women’s action. Disney+ will also carry the Final Four. Cable providers, then, are required for fans hoping to seamlessly hop back and forth between the two brackets, even as fans shift to a streaming-first future.
That last sentence is the key: Warner Bros. Discovery only has access to the games it owns rights to; YouTube TV, by virtue of being a virtual Multichannel Video Programming Distributor (vMVPD), has access to every game that is on cable, which is all of them. That lets the service offer an objectively better multiview experience.
YouTube TV’s Virtual Advantage
Multiview isn’t a new idea; in 1983 George Schnurle III invented the MultiVision:
Mrmazda, CC-SA
This image is of the MultiVision 1.1, which took in four composite inputs; the 3.1 model included two built-in tuners — you provided the antenna. The Multivision didn’t provide multiview a la YouTube TV, but rather picture-in-picture, support for which was eventually built into TVs directly.
Picture-in-picture, however, assumed that consumers had easy access to TV signals; this was a reasonable assumption when signals came in over-the-air or via basic cable. That changed in the late 1990s with the shift to digital cable, which required a set-top box to decrypt; most TVs only had one, and the picture-in-picture feature faded away. This loss was made up in part by the addition of DVR functionality to most of those set-top boxes; with time-shifting you couldn’t watch two things at once, but you could watch two things that aired at the same time.
Cable companies offered DVR functionality in response to the popularity of TiVo; when the first model launched in 1999 it too relied on the relative openness of TV signals. Later models needed cable cards, which were mandated by the FCC in 2007; that mandate was repealed in 2020, as the good-enough nature of cable set-top boxes effectively killed the market for TiVo and other 3rd-party tuners.
The first vMVPD, meanwhile, was Sling TV, which launched in 2015.2 YouTube TV launched two years later, with an old Google trick: unlimited storage for your cloud DVR, which you could watch anywhere in the U.S. on any device. That was possible because the point of integration for YouTube TV, unlike traditional cable, was on Google’s servers, not a set-top box (which itself was a manifestation of traditional MVPD’s point of integration being the cable into your house).
This point of integration also explains why it was YouTube TV that came up with the modern implementation of multiview: Google could create this new feature centrally and make it available to everyone without needing to install high-powered set-top boxes in people’s homes. Indeed, this explains one of the shortcomings of multiview: because Google can not rely on viewers having high powered devices capable of showing four independent video streams, Google actually pre-mixes the streams into a single video feed on their servers.
YouTube TV + NFL Sunday Ticket
I mentioned above that YouTube TV offered multiview for March Madness starting last year, but that’s not quite right: a subset of the consumer base actually got access for March Madness in 2023; that was a beta test for the real launch, which was the 2023 NFL season. That was the first year that Google had the rights to NFL Sunday Ticket, which lets subscribers view out-of-market games. NFL Sunday Ticket was a prerequisite for multiview, because without it you would have access to at most two football games at a time; once you could watch all of the games, the utility was obvious.
The point of this Article is not multiview; it’s a niche use case for events like March Madness or football fanatics on Sunday afternoons. What is notable about the latter example, however, is that Google needed to first secure the rights to NFL Sunday Ticket. This, unlike March Madness, wasn’t a situation where every game was already on cable, and thus accessible to YouTube TV; Google needed to pay $2 billion/year to secure the necessary rights to make multiview work.
That’s a high price, even if multiview is cool; it seems unlikely that Google will ever make its money back directly. That, though, is often the case with the NFL. Back in 1993 Rupert Murdoch shocked the world by buying NFL broadcasting rights for a then-unprecedented $395 million/year, $100 million/year more than CBS was offering for the same package. Sports Illustrated explained his reasoning:
There are skeptics who think that Murdoch will lose his custom-made shirt over the NFL deal; one estimate has him losing $500 million over the next four years. Says Murdoch, “I’ve seen those outrageous numbers. We’ll lose a few million in the first year, but even if it was 40 or 50 million, it would be tax deductible. It was a cheap way of buying a network.”
What Murdoch meant was that demand for the NFL — which had already built ESPN — would get Fox into the cities where it didn’t yet exist, and improve its affiliate station’s standing (many of which Murdoch owned) in cities where they were weak and buried on the inferior UHF band. And, of course, that is exactly what happened.
NFL Sunday Ticket is not, to be sure, the same as regular NFL rights; it is much more of a niche product with a subscription business model. That, though, is actually a good thing from Google’s perspective: the company’s opportunity is not to build a TV station, but rather a TV Aggregator.
YouTube TV’s Aggregation Potential
Google announced the NFL deal a month after it launched Primetime Channels, a marketplace for streaming services along the lines of Amazon’s Prime Video Channels or Apple TV Channels; I wrote in early 2023:
The missing piece has been — in contrast to Apple and Amazon in particular — other streaming services. Primetime Channels, though, is clearly an attempt to build up YouTube’s own alternative to the Apple TV App Store or Amazon Prime Video Marketplace. This, as I noted last month, is why I think YouTube’s extravagant investment in NFL Sunday Ticket makes sense: it is a statement of intent and commitment that the service wants to use to convince other streaming services to come on board. The idealized future is one where YouTube is the front-door of all video period, whether that be streaming, linear, or user-generated.
YouTube’s big advantage, as I noted in that Update, is that it has exclusive access to YouTube content; it is the only service that can offer basically anything you might want to watch on TV:
YouTube TV has linear television, which remains important for sports
YouTube proper dominates user-generated content
Primetime Channels is a way to bring other streaming services on board
The real potential with streaming channels, however, is to go beyond selling subscriptions on an ad-hoc basis and actually integrating them into a single interface to drive discoverability and on-demand conversions. How useful would it be to see everything that is on in one place, and be able to either watch with one click, or subscribe with two?
This is going to be an increasingly pressing need as sports in particular move to streaming. It used to be that all of the sports you might watch were in a centralized place: the channel guide on your set-top box. Today, however, many sports are buried in apps. Prominent examples include Amazon Thursday Night Football and Peacock’s exclusive NFL playoff games, but as a Wisconsin fan I’ve already experienced the challenge of an increasing number of college basketball games being exclusively streamed on Peacock; the problem is only going to get worse next season when an increasing number of NBA games are on Amazon and Peacock, and when ESPN releases a standalone streaming app with all of its games.
The challenge for any one of these services is the same one seen with Max’s multiview offering: any particular streaming service is limited to its own content. Sure, any one of these services could try and build this offering anyways — ESPN is reportedly considering it — but then they run into the problem of not being a platform or marketplace with a massive audience already in place.
The reason why that is an essential prerequisite is that executing on this vision will require forming partnerships with all of the various streamers — or at least those with live events like sports. On one hand, of course each individual streamer wants to own the customer relationship; on the other hand, sports rights both cost a lot of money and also lose their value the moment an event happens. That means they are motivated to trade away customer control and a commission for more subscribers, which works to the benefit of whoever can marshal the most demand, and YouTube, thanks primarily to its user-generated content, has the largest audience of all, and thanks to YouTube TV, is the only service that can actually offer everything.
Google’s Product Problem
Two quick questions for the audience:
Did you know that Primetime Channels existed?
How do you subscribe to Primetime Channels?
The answer to number 2 is convoluted, to say the least; on a PC, you click the hamburger button in the upper left, then click “Your movies & TV”, then click the “Browse” tab, and there you will finally find Primetime channels; on mobile the “Your movies & TV” is found by clicking your profile photo on the bottom right.
And, once you finally figure this out, you see a pretty pathetic list:
As the arrow indicates, there are more options, but the only one of prominence is Paramount+; there is no Disney+, Peacock, Amazon Prime Video, Apple TV+, or Netflix.
Netflix’s resistance to being aggregated is long-running; they were the stick in the mud when Apple tried to aggregate streaming a decade ago. The company gets away with it — and are right to resist — because they have the largest user base amongst subscription platforms. The biggest bull case for Netflix is that many of the other streamers throw in the towel and realize they are better off just selling content to Netflix.
Disney+ actually could pull off a fair bit of what YouTube is primed to do: no, Disney doesn’t have YouTube’s user-generated content, but the company does have Hulu Live, which gives a potential Aggregation offering access to content still on linear TV.
Amazon and Apple are Google’s most obvious competitors when it comes to building an Aggregator for streaming services, and they have the advantage of owning hardware to facilitate transactions.
That leaves Peacock, and this is where I hold Google responsible. Peacock has large bills and a relatively small userbase; there is also a Peacock app for both Amazon devices (although you have to subscribe to Peacock directly) and Apple devices (where Apple enforces an in-app subscription offering). If Google is serious about Primetime Channels specifically, and being a streaming and sports Aggregator generally, then it should have Peacock available as an offering.
That’s the thing, though: it’s not clear that Google has made any sort of progress in achieving the vision I perceived two years ago in the wake of the launch of Primetime Channels and the NFL Sunday Ticket deal. Yes, YouTube continues to grow, particularly on TVs, and yes, multivision is slowly getting better, but both of those are products of inertia; is Google so arthritic that it can’t make a play to dominate an entertainment industry that is getting religion about the need to acquire and keep customers profitably? That’s exactly why Aggregators gain power over suppliers: they solve their demand problem. And yet Primetime Channels might as well not even exist, given how buried it is, and it might as well be, given that Google hasn’t signed a meaningful new deal since launch.
Google’s Wiz Acquisition
This is all convoluted way to explain why I approve of Google’s decision to pay $32 billion in cash for Wiz, a cybersecurity firm that has absolutely nothing to do with the future of TV. From Bloomberg:
Google parent Alphabet Inc. agreed to acquire cybersecurity firm Wiz Inc. for $32 billion in cash, reaching a deal less than a year after initial negotiations fell apart because the cloud-computing startup wanted to stay independent. Wiz will join the Google Cloud business once the deal closes, the companies said in a statement on Tuesday. The takeover is subject to regulatory approvals and is likely to close next year, they said.
The deal, which would be Alphabet’s largest to date, comes after Wiz turned down a $23 billion bid from the internet search leader last year after several months of discussions. At the time, Wiz walked away after deciding it could ultimately be worth more by pursuing an initial public offering company. Concerns about regulatory challenges also influenced the decision. The companies have agreed to a breakup fee of about 10% of the deal value, or $3.2 billion, if the deal doesn’t close, according to a person familiar with the matter. Shares of Alphabet fell nearly 3% in New York on Tuesday.
Wiz provides cybersecurity solutions for multi-cloud environments, and is growing fast. This makes it a natural fit for Google Cloud, which is a distant third place to AWS and Microsoft Azure. Google Cloud’s biggest opportunity for growth is to be a service that is used in addition to a large corporation’s existing cloud infrastructure, and Wiz provides both a beachhead into those organizations and also a solution to managing a multi-cloud setup.
Google Cloud’s selling point — the reason it might expand beyond a Wiz beachhead — are Google’s AI offerings. Google continues to have excellent AI research and the best AI infrastructure; where the company is struggling is product, particularly in the consumer space, thanks to some combination of fear of disruption and, well, the fact that product capability seems to be the first casualty of a monopoly (Apple’s declining product chops, particularly in software and obviously AI, is another example).
The company’s tortoise-like approach to TV lends credence to the latter explanation: Google is in an amazing position in TV, thanks to the long-ago acquisition of YouTube and the launch of YouTube TV, but it has accomplished little since then beyond agreeing to pay the NFL a lot of money. Arguably the ideal solution to this sort of malaise, at least from a shareholder perspective, would be to simply collect monopoly rents and return the money to shareholders at a much higher rate than Google has to date; absent that, buying product innovation seems like the best way to actually accomplish anything.
In other words, while I understand the theory of people who think that Google ought to just build Wiz’s functionality instead of paying a huge revenue multiple for a still-unprofitable startup, I think the reality of a company like Google is that said theory would run into the morass that is product development in a monopoly. It simply would not ship, and would suck if it did. Might as well pay up for momentum in a market that has some hope of leveraging the still considerable strengths that exist beneath the flab.
Technically 68; there are four games on Tuesday that trim the field to 64 ↩
The original Sling TV was a cable card device that allowed you to watch your TV from anywhere in the world; it was massively popular amongst expats here in Taiwan ↩
It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of Light, it was the season of Darkness, it was the spring of hope, it was the winter of despair, we had everything before us, we had nothing before us, we were all going direct to Heaven, we were all going direct the other way — in short, the period was so far like the present period that some of its noisiest authorities insisted on its being received, for good or for evil, in the superlative degree of comparison only. — Charles Dickens, A Tale of Two Cities
Apple’s Bad Week
Apple has had the worst of weeks when it comes to AI. Consider this commercial which the company was running incessantly last fall:
In case you missed the fine print in the commercial, it reads:
Apple Intelligence coming fall 2024 with Siri and device language set to U.S. English. Some features and languages will be coming over the next year.
“Next year” is doing a lot of work, now that the specific feature detailed in this commercial — Siri’s ability to glean information from sources like your calendar — is officially delayed. Here is the statement Apple gave to John Gruber at Daring Fireball:
Siri helps our users find what they need and get things done quickly, and in just the past six months, we’ve made Siri more conversational, introduced new features like type to Siri and product knowledge, and added an integration with ChatGPT. We’ve also been working on a more personalized Siri, giving it more awareness of your personal context, as well as the ability to take action for you within and across your apps. It’s going to take us longer than we thought to deliver on these features and we anticipate rolling them out in the coming year.
It was a pretty big surprise, even at the time, that Apple, a company renowned for its secrecy, was so heavily advertising features that did not yet exist; I also, in full disclosure, thought it was all an excellent idea. From my post-WWDC Update:
The key part here is the “understanding personal context” bit: Apple Intelligence will know more about you than any other AI, because your phone knows more about you than any other device (and knows what you are looking at whenever you invoke Apple Intelligence); this, by extension, explains why the infrastructure and privacy parts are so important.
What this means is that Apple Intelligence is by-and-large focused on specific use cases where that knowledge is useful; that means the problem space that Apple Intelligence is trying to solve is constrained and grounded — both figuratively and literally — in areas where it is much less likely that the AI screws up. In other words, Apple is addressing a space that is very useful, that only they can address, and which also happens to be “safe” in terms of reputation risk. Honestly, it almost seems unfair — or, to put it another way, it speaks to what a massive advantage there is for a trusted platform. Apple gets to solve real problems in meaningful ways with low risk, and that’s exactly what they are doing.
Contrast this to what OpenAI is trying to accomplish with its GPT models, or Google with Gemini, or Anthropic with Claude: those large language models are trying to incorporate all of the available public knowledge to know everything; it’s a dramatically larger and more difficult problem space, which is why they get stuff wrong. There is also a lot of stuff that they don’t know because that information is locked away — like all of the information on an iPhone. That’s not to say these models aren’t useful: they are far more capable and knowledgable than what Apple is trying to build for anything that does not rely on personal context; they are also all trying to achieve the same things.
So is Apple more incompetent than these companies, or was my evaluation of the problem space incorrect? Much of the commentary this week assumes point one, but as Simon Willison notes, you shouldn’t discount point two:
I have a hunch that this delay might relate to security. These new Apple Intelligence features involve Siri responding to requests to access information in applications and then performing actions on the user’s behalf. This is the worst possible combination for prompt injection attacks! Any time an LLM-based system has access to private data, tools it can call, and exposure to potentially malicious instructions (like emails and text messages from untrusted strangers) there’s a significant risk that an attacker might subvert those tools and use them to damage or exfiltrating a user’s data.
Willison links to a previous piece of his on the risk of prompt injections; to summarize the problem, if your on-device LLM is parsing your emails, what happens if one of those emails contains malicious text perfectly tuned to make your on-device AI do something you don’t want it to? We intuitively get why code injections are bad news; LLMs expand the attack surface to text generally; Apple Intelligence, by being deeply interwoven into the system, expands the attack surface to your entire device, and all of that precious content it has unique access to.
Needless to say, I regret not raising this point last June, but I’m sure my regret pales in comparison to Apple executives and whoever had to go on YouTube to pull that commercial over the weekend.
Apple’s Great Week
Apple has had the best of weeks when it comes to AI. Consider their new hardware announcements, particularly the Mac Studio and its available M3 Ultra; from the company’s press release:
Apple today announced M3 Ultra, the highest-performing chip it has ever created, offering the most powerful CPU and GPU in a Mac, double the Neural Engine cores, and the most unified memory ever in a personal computer. M3 Ultra also features Thunderbolt 5 with more than 2x the bandwidth per port for faster connectivity and robust expansion. M3 Ultra is built using Apple’s innovative UltraFusion packaging architecture, which links two M3 Max dies over 10,000 high-speed connections that offer low latency and high bandwidth. This allows the system to treat the combined dies as a single, unified chip for massive performance while maintaining Apple’s industry-leading power efficiency. UltraFusion brings together a total of 184 billion transistors to take the industry-leading capabilities of the new Mac Studio to new heights.
“M3 Ultra is the pinnacle of our scalable system-on-a-chip architecture, aimed specifically at users who run the most heavily threaded and bandwidth-intensive applications,” said Johny Srouji, Apple’s senior vice president of Hardware Technologies. “Thanks to its 32-core CPU, massive GPU, support for the most unified memory ever in a personal computer, Thunderbolt 5 connectivity, and industry-leading power efficiency, there’s no other chip like M3 Ultra.”
That Apple released a new Ultra chip wasn’t a shock, given there was an M1 Ultra and M2 Ultra; almost everything about this specific announcement, however, was a surprise.
Start with the naming. Apple chip names have two components: M_ refers to the core type, and the suffix to the configuration of those cores. Therefore, to use the M1 series of chips as an example:
Perf Cores
Efficiency Cores
GPU Cores
Max RAM
Bandwidth
M1
4
4
8
16GB
70 GB/s
M1 Pro
8
4
16
32GB
200 GB/s
M1 Max
8
2
32
64GB
400 GB/s
M1 Ultra
16
4
64
128GB
800 GB/s
The “M1” cores in question were the “Firestorm” high-performance core, “Icestorm” energy-efficient core, and a not-publicly-named GPU core; all three of these cores debuted first on the A14 Bionic chip, which shipped in the iPhone 12.
The suffix, meanwhile, referred to some combination of increased core count (both CPU and GPU), as well as an increased number of memory controllers and associated bandwidth (and, in the case of the M1 series, faster RAM). The Ultra, notably, was simply two Max chips fused together; that’s why all of the numbers simply double.
The M2 was broadly similar to the M1, at least in terms of the relative performance of the different suffixes. The M2 Ultra, for example, simply doubled up the M2 Max. The M3 Ultra, however, is unique when it comes to max RAM:
Perf Cores
Efficiency Cores
GPU Cores
Controllers
Max RAM
Bandwidth
M3
4
4
10
8
32GB
100 GB/S
M3 Pro
6
6
18
12
48GB
150 GB/s
M3 Max
12
4
40
32
128GB
400 GB/s
M3 Ultra
24
8
80
64
512GB
800 GB/s
I can’t completely vouch for every number on this table (which was sourced from Wikipedia), as Apple hasn’t yet released the full technical details of the M3 Ultra, and it’s not yet available for testing. What seems likely, however, is that instead of simply doubling up the M3 Max, Apple also reworked the memory controllers to address double the memory. That also explains why the M3 Ultra came out so much later than the rest of the family — indeed, the Mac Studio base chip is actually the M4 Max.
The wait was worth it, however: what makes Apple’s chip architecture unique is that that RAM is shared by the CPU and GPU, and not in the carve-out way like integrated graphics of old; rather, every part of the chip — including the Neural Processing Units, which I didn’t include on these tables — has full access to (almost1) all of the memory all of the time.
What that means in practical terms is that Apple just shipped the best consumer-grade AI computer ever. A Mac Studio with an M3 Ultra chip and 512GB RAM can run a 4-bit quantized version of DeepSeek R1 — a state-of-the-art open-source reasoning model — right on your desktop. It’s not perfect — quantization reduces precision, and the memory bandwidth is a bottleneck that limits performance — but this is something you simply can’t do with a standalone Nvidia chip, pro or consumer. The former can, of course, be interconnected, giving you superior performance, but that costs hundreds of thousands of dollars all-in; the only real alternative for home use would be a server CPU and gobs of RAM, but that’s even slower, and you have to put it together yourself.
Apple didn’t, of course, explicitly design the M3 Ultra for R1; the architectural decisions undergirding this chip were surely made years ago. In fact, if you want to include the critical decision to pursue a unified memory architecture, then your timeline has to extend back to the late 2000s, whenever the key architectural decisions were made for Apple’s first A4 chip, which debuted in the original iPad in 2010.
Regardless, the fact of the matter is that you can make a strong case that Apple is the best consumer hardware company in AI, and this week affirmed that reality.
Apple Intelligence vs. Apple Silicon
It’s probably a coincidence that the delay in Apple Intelligence and the release of the M3 Ultra happened in the same week, but it’s worth comparing and contrasting why one looks foolish and one looks wise.
Apple Silicon
Start with the latter: Tony Fadell told me the origin story of Apple Silicon in a 2022 Stratechery Interview; the context of the following quote was his effusive praise for Samsung, which made the chips for the iPod and the first several models of the iPhone:
Samsung was an incredible partner. Even though they got sued, they were an incredible partner, they had to exist for the iPod to be as successful and for the iPhone to even exist. That happened. During that time, obviously Samsung was rising up in terms of its smartphones and Android and all that stuff, and that’s where things fell apart.
At the same time, there was the strategic thing going on with Intel versus ARM in the iPad, and then ultimately iPhone where there’s that fractious showdown that I had with various people at Apple, including Steve, which was Steve wanted to go Intel for the iPad and ultimately the iPhone because that’s the way we went with the Mac and that was successful. And I was saying, “No, no, no, no! Absolutely not!” And I was screaming about it and that’s when Steve was, well after Intel lost the challenge, that’s when Steve was like, “Well, we’re going to go do our own ARM.” And that’s where we bought P.A. Semi.
So there was the Samsung thing happening, the Intel thing happening, and then it’s like we need to be the master of our own destiny. We can’t just have Samsung supplying our processors because they’re going to end up in their products. Intel can’t deliver low power embedded the way we would need it and have the culture of quick turns, they were much more standard product and non custom products and then we also have this, “We got to have our own strategy to best everyone”. So all of those things came together to make what happened happen to then ultimately say we need somebody like TSMC to build more and more of our chips. I just want to say, never any of these things are independently decisions, they were all these things tied together for that to pop out of the oven, so to speak.
This is such a humbling story for me as a strategy analyst; I’d like to spin up this marvelous narrative about Apple’s foresight with Apple Silicon, but like so many things in business, it turns out the best consumer AI chips were born out of pragmatic realities like Intel not being competitive in mobile, and Samsung becoming a smartphone competitor.
Ultimately, though, the effort is characterized by four critical qualities:
Time: Apple has been working on Apple Silicon for 17 years.
Motivation: Apple was motivated to build Apple Silicon because having competitive and differentiated mobile chips was deemed essential to their business.
Differentiation: Apple’s differentiation has always been rooted in the integration of hardware and software, and controlling their own chips let them do exactly that, wringing out unprecedented efficiency in particular.
Iteration: The M3 Ultra isn’t Apple’s first chip; it’s not even the first M chip; heck, it’s not even the first M3! It’s the result of 17 years of iteration and experimentation.
Apple Intelligence
Notice how these qualities differ when it comes to Apple Intelligence:
Time: The number one phrase that has been used to characterize Apple’s response to the ChatGPT moment in November 2022 is flat-footed, and that matches what I have heard anecdotally. That, by extension, means that Apple has been working on Apple Intelligence for at most 28 months, and that is almost certainly generous, given that the company likely took a good amount of time to figure out what its approach would be. That not nothing — xAI went from company formation to Grok 3 in 19 months — but it’s certainly not 17 years!
Motivation: If you look at Apple’s earnings calls in the wake of ChatGPT, February 2023, May 2023, and August 2023, all contain some variation of “AI and machine learning have been integrated into our products for years, and we’ll continue to be thoughtful about how we implement them”; finally in November 2023 CEO Tim Cook said the company was working on something new:
In terms of generative AI, we have — obviously, we have work going on. I’m not going to get into details about what it is, because, as you know, we don’t — we really don’t do that. But you can bet that we’re investing, we’re investing quite a bit, we’re going to do it responsibly and it will — you will see product advancements over time that where the — those technologies are at the heart of them.
First, this obviously has bearing on the “time” point above; secondly, one certainly gets the sense that Apple, after tons of industry hype and incessant questions from analysts, very much representing the concerns of shareholders, felt like they had no choice but to be doing something with generative AI. In other words — and yes, this is very much driving with the rearview mirror — Apple didn’t seem to be working on generative AI because they felt it was essential to their product vision, but rather because they had to keep up with what everyone else was doing.
Differentiation: This is the most alluring part of the Apple Intelligence vision, which I myself hyped up from the beginning: Apple’s exclusive access to its users’ private information. What is interesting to consider, however, beyond the security implications, is the difference between “exclusivity” and “integration”.
Consider your address book: the iOS SDK included the Contacts API, which gave any app on the system full access to your contacts without requiring explicit user permission. This was essential to the early success of services like WhatsApp, which cleverly bootstrapped your network by using phone numbers as unique IDs; this meant that pre-existing username-based networks like Skype and AIM were actually at a disadvantage on iOS. iMessage did the same thing when it launched in 2011, and then Apple started requiring user permission to access your contacts in 2012.
Even this amount of access, however, paled in comparison to the Mac, where developers could access information from anywhere on the system. iOS, on the other hand, put apps in sandboxes, cut off from other apps and system information outside of APIs like the Contacts API, all of which have become more and more restricted over time. Apple made these decisions for very good reasons, to be clear: iOS is a much safer and secure environment than macOS; increased restrictions generally mean increased privacy, albeit at the cost of decreased competition.
Still, it’s worth pointing out that exclusive access to data is downstream of a policy choice to exclude third parties; this is distinct from the sort of hardware and software integration that Apple can exclusively deliver in the pursuit of superior performance. This distinction is subtle, to be sure, but I think it’s notable that Apple Silicon’s differentiation was in the service of building a competitive moat, while Apple Intelligence’s differentiation was about maintaining one.
Iteration: From one perspective, Apple Intelligence is the opposite of an evolved system: Apple put together an entire suite of generative AI capabilities, and aimed to launch them all in iOS 18. Some of these, like text manipulation and message summaries, were straightforward and made it out the door without a problem; others, particularly the reimagined Siri and its integration with 3rd party apps and your personal data, are now delayed. It appears Apple tried to do too much all at once.
The Incumbent Advantage
At the same time, it’s not as if Siri is new; the voice assistant launched in 2011, alongside iMessage. In fact, though, Siri has always tried to do too much too soon; I wrote last week about the differences between Siri and Alexa, and how Amazon was wise to focus their product development on the basics — speed and accuracy — while making Alexa “dumber” than Siri tried to be, particularly in its insistence on precise wording instead of attempting to figure out what you meant.
To that end, this speaks to how Apple could have been more conservative in its generative AI approach (and, I fear, Amazon too, given my skepticism of Alexa+): simply make a Siri that works. The fact of the matter is that Siri has always struggled with delivering on its promised functionality, but a lot of its shortcomings could have been solved by generative AI. Apple, however, promised much more than this at last year’s WWDC: Siri wasn’t simply going to work better, it was actually going to understand and integrate your personal data and 3rd-party apps in a way that had never been done before.
Again, I applauded this at the time, so this is very much Monday-morning quarterbacking. I increasingly suspect, however, we are seeing a symptom of big-company disease that I hadn’t previously considered: while one failure state in the face of new technology is moving too slowly, the opposite failure state is assuming you can do too much too quickly, when simply delivering the basics would be more than good enough.
Consider home automation: the big three players in the space are Siri and Alexa and Google Assistant. What makes these companies important is not simply that they have devices you can put in your home and talk to, but also that there is an entire ecosystem of products which work with them. Given that, consider two possible products in the space:
OpenAI releases a ChatGPT speaker that you can talk to and interact with; it works brilliantly and controls, well, it doesn’t control anything, because the ecosystem hasn’t adopted it. OpenAI would need to work diligently to build out partnerships with everyone from curtain makers to smart light to locks and more; that’s hard enough in its own right, and even more difficult when you consider that many of these objects are only installed once and updated rarely.
Apple or Amazon or Google update their voice assistants with basic LLMs. Now, instead of needing to use precise language, you can just say whatever you want, and the assistant can figure it out, along with all of the other LLM niceties like asking about random factoids.
In this scenario the Apple/Amazon/Google assistants are superior, even if their underlying LLMs are worse, or less capable than OpenAI’s offering, because what the companies are selling is not a standalone product but an ecosystem. That’s the benefit of being a big incumbent company: you have other advantages you can draw on beyond your product chops.
What is striking about new Siri — and, I worry, Alexa+ — is the extent to which they are focused on being compelling products in their own right. It’s very clever for Siri to remember who I had coffee with; it’s very useful — and probably much more doable — to reliably turn my lights on and off. Apple (and I suspect Amazon) should have absolutely nailed the latter before promising to deliver the former.
If you want to be generous to Apple you could make the case that this was what they were trying to deliver with the Siri Intents expansion: developers could already expose parts of their apps to Siri for things like music playback, and new Siri was to build on that framework to enhance its knowledge about a user’s context to provide useful answers. This, though, put Apple firmly in control of the interaction layer, diminishing and commoditizing apps; that’s what an Aggregator does, but what if Apple went in a different direction?
There is certainly an argument to be made that these two philosophies arise out of their historical context; it is no accident that Apple and Microsoft, the two “bicycle of the mind” companies, were founded only a year apart, and for decades had broadly similar business models: sure, Microsoft licensed software, while Apple sold software-differentiated hardware, but both were and are at their core personal computer companies and, by extension, platforms.
Google and Facebook, on the other hand, are products of the Internet, and the Internet leads not to platforms but to Aggregators. While platforms need 3rd parties to make them useful and build their moat through the creation of ecosystems, Aggregators attract end users by virtue of their inherent usefulness and, over time, leave suppliers no choice but to follow the Aggregators’ dictates if they wish to reach end users.
The business model follows from these fundamental differences: a platform provider has no room for ads, because the primary function of a platform is to provide a stage for the applications that users actually need to shine. Aggregators, on the other hand, particularly Google and Facebook, deal in information, and ads are simply another type of information. Moreover, because the critical point of differentiation for Aggregators is the number of users on their platform, advertising is the only possible business model; there is no more important feature when it comes to widespread adoption than being “free.”
Still, that doesn’t make the two philosophies any less real: Google and Facebook have always been predicated on doing things for the user, just as Microsoft and Apple have been built on enabling users and developers to make things completely unforeseen.
I said this was romantic, but the reality of Apple’s relationship with developers, particularly over the last few years as the growth of the iPhone has slowed, has been considerably more antagonistic. Apple gives lip service to the role developers played in making the iPhone a compelling platform — and in collectively forming a moat for iOS and Android — but its actions suggest that Apple views developers as a commodity: necessary in aggregate, but mostly a pain in the ass individually.
This is all very unfortunate, because Apple — in conjunction with its developers — is being presented with an incredible opportunity by AI, and it’s one that takes them back to their roots: to be a platform.
Start with the hardware: while the M3 Ultra is the biggest beast on the block, all of Apple’s M chips are highly capable, particularly if you have plenty of RAM. I happen to have an M2 MacBook Pro with 96GB of memory (I maxed out for this specific use case), which lets me run Mixtral 8x22B, an open-source model from Mistral with 141 billion parameters, at 4-bit quantization; I asked it a few questions:
You don’t need to actually try and read the screen-clipping; the output is pretty good, albeit not nearly as detailed and compelling as what you might expect from a frontier model. What’s amazing is that it exists at all: that answer was produced on my computer with my M2 chip, not in the cloud on an Nvidia datacenter GPU. I didn’t need to pay a subscription, or worry about rate limits. It’s my model on my device.
What’s arguably even more impressive is seeing models run on your iPhone:
This is a much smaller model, and correspondingly less capable, but the fact it is running locally on a phone is amazing!
Apple is doing the same thing with the models that undergird Apple Intelligence — some models run on your device, and others on Apple’s Private Cloud Compute — but those models aren’t directly accessible by developers; Apple only exposes writing tools, image playground, and Genmoji. And, of course, they ask for your app’s data for Siri, so they can be the AI Aggregator. If a developer wants to do something unique, they need to bring their own model, which is not only very large, but hard to optimize for a specific device.
What Apple should do instead is make its models — both local and in Private Cloud Compute — fully accessible to developers to make whatever they want. Don’t limit them to cutesy-yet-annoying frameworks like Genmoji or sanitized-yet-buggy image generators, and don’t assume that the only entity that can create something compelling using developer data is the developer of Siri; instead return to the romanticism of platforms: enabling users and developers to make things completely unforeseen. This is something only Apple could do, and, frankly, it’s something the entire AI industry needs.
When the M1 chip was released I wrote an Article called Apple’s Shifting Differentiation. It explained that while Apple had always been about the integration of hardware and software, the company’s locus of differentiation had shifted over time:
When OS X first came out, Apple’s differentiation was software: Apple hardware was stuck on PowerPC chips, woefully behind Intel’s best offerings, but developers in particular were lured by OS X’s beautiful UI and Unix underpinnings.
When Apple moved to Intel chips, its hardware was just as fast as Windows hardware, allowing its software differentiation to truly shine.
Over time, as more and more applications moved to the web, the software differences came to matter less and less; that’s why the M1 chip was important for the Mac’s future.
Apple has the opportunity with AI to press its hardware advantage: because Apple controls the entire device, they can guarantee to developers the presence of particular models at a particular level of performance, backed by Private Cloud Compute; this, by extension, would encourage developers to experiment and build new kinds of applications that only run on Apple devices.
This doesn’t necessarily preclude finally getting new Siri to work; the opportunity Apple is pursuing continues to make sense. At the same time, the implication of the company’s differentiation shifting to hardware is that the most important job for Apple’s software is to get out of the way; to use Apple’s history as analogy, Siri is the PowerPC of Apple’s AI efforts, but this is a self-imposed shortcoming. Apple is uniquely positioned to not do everything itself; instead of seeing developers as the enemy, Apple should deputize them and equip them in a way no one else in technology can.
