Apple and the Power of Platforms, AI Comes for Google Search, ChatGPT Comes for Higher Education
Summary
Apple’s iPhone user base—not developer goodwill—is the App Store moat, leaving government intervention as the only credible constraint on its economics. Ben Thompson rejects the usual chicken-and-egg framing: “The reality is users come first,” and valuable users make developers tolerate almost any obstacle. His categorical call: developers will not flee, but Apple’s conduct can invite regulation or an adverse judicial decision.
Thompson says the App Store’s theoretical market-clearing rate is either nearly everything or nothing, making 30% an arbitrary exercise of platform power. Apple could demand “99.9” because developers cannot abandon iPhone users, while a prohibition on charging would still leave Apple supplying APIs because apps are indispensable to the device. He nevertheless defends commissions for Apple’s integrated in-app purchase flow; the abuse is preventing developers from telling users about outside commerce, not merely requiring them to use their own web checkout.
Apple has optimized itself into a brittle dependence on Google’s roughly $20 billion to $25 billion in annual search payments. Thompson describes the money as pure profit and frames it as perhaps a quarter of Apple’s profit, making Eddy Cue’s warnings about AI search strategically conflicted: Apple needs to demonstrate competition without losing the cash. His broader Cook critique is that Apple has “squeezed the juice out of every aspect of their business” across Google, China, and the App Store at the expense of resilience.
The first reported decline in Safari search volume in two decades is an unusually concrete signal that generative AI is taking activity from Google, not merely adding another app category. Thompson calls AI usage a “one-way ratchet” closer to the personal computer or phone than to Instagram: serious users tend to use it more, not burn out and leave. Google still controls high-value commercial queries, but a shift in Safari options or defaults could accelerate the change if Apple adds ChatGPT, Perplexity, or Anthropic options.
Google Search faces real product substitution, but its advertising machinery may endure far longer than the bearish narrative implies. Thompson cites commercial-query dominance, rising CPMs, difficult-to-replicate ad infrastructure, and Google’s claim that AI Overviews monetize as well as conventional search. “These things hold on longer than you think,” while new products often layer on top rather than fully replace incumbents.
ChatGPT is exposing higher education’s weak underlying economics more than it is destroying a previously healthy institution. Thompson’s blunt thesis is that “AI is not ruining college. College was ruined,” with degrees functioning mainly as filters for conscientiousness and basic intelligence after employers outsourced credentialing to universities. The wage premium remains large enough that he still considers sending children to college obvious, even as credentials become costlier and less connected to learning.
AI may divide the workforce between high-agency people whose output is dramatically amplified and a much larger population that delegates effort while consuming endless entertainment. Andrew Sharp insists that struggling through writing builds critical thought; Thompson concedes the concern but asks whether many students ever sought learning rather than the credential. The darkest scenario is economically productive in aggregate yet socially hollow: a few people gain organization-scale leverage while many become “vaguely dissatisfied” but comfortably distracted.
Deep dive
1. Microsoft taught Thompson that distribution beats developer persuasion
Thompson entered Microsoft deeply skeptical of Windows 8, having deleted critical tweets when employment made “discretion the better part of valor.” He nevertheless gave the project “the good college try” and quickly moved onto the Windows App Store, where his nominally level-61 position carried responsibilities closer to level 64.
His developer-relations work reached Microsoft’s largest partners and its worldwide “scale motion.” A presentation he rebuilt from scratch was eventually used by thousands of field employees; he also brought Kindle and Automattic’s WordPress app into the store’s initial featured lineup while learning the publishing industry’s operational constraints.
The prelaunch pitch was potential: Windows had a billion users, so developers should get ahead of the curve as they wished they had on iPhone or Android. After Windows 8 shipped, “there was no curve to get ahead of”—developers had gone ahead only to become “stuck in the desert.”
That failure taught Thompson that launching an app is not merely a funded development project. Maintenance persists, unstable APIs create genuine delivery risk, and even a small user base can become vocal enough to damage a company’s broader reputation; the job shifted from selling possibility to “managing dashed expectations.”
2. Users create platforms, and applications follow demand
Thompson’s central correction to the platform chicken-and-egg story is categorical: “The reality is users come first.” A compelling product accumulates users, those users create demand for applications, and developers arrive even when platform policies are hostile because reaching valuable customers matters more than platform goodwill.
The iPhone therefore gained a complete app ecosystem despite Apple’s “egregious” restrictions. Thompson acknowledged that he once overstated developer leverage, but now sees the App Store conflict as “just a fight about money”: there is basically nothing Apple can do that would make important developers abandon billions of valuable users.
Vision Pro demonstrates the boundary. Before launch, poor developer relations mattered because Apple was selling an unproven future—“I’ve stabbed you in the back for 10 years. Will you scratch my back?” Once a platform has millions of users, however, Thompson expects Netflix or any comparable service to build for it regardless of accumulated resentment.
He also rejects the standard Windows-versus-Mac parable: Windows won through DOS and IBM distribution before the Mac arrived, not because modularity later disrupted integration. That mistaken history fed Clayton Christensen’s analysis and the broader myth that recruiting developers creates platforms; Thompson’s inversion is that successful products acquire developers “sort of by default.”
3. Platform infrastructure deserves latitude—and focused antitrust
Thompson defines a platform as an operating-system-like foundation with APIs on which third parties depend: Photoshop requires macOS, Windows, or iPadOS to exist. Platforms generate enormous downstream value and deserve substantial operating latitude, but their infrastructural position also grants “basically untrammeled power.”
Aggregators are different because users can leave: Google’s claim that another service is “a click away” is substantially true. Platforms resemble railroads—fixed infrastructure governing market access—while aggregators must repeatedly earn consumer choice, even when billions of people happen to make the same choice.
That distinction drives Thompson’s frustration with anti-monopolists who prioritized Google and Facebook while overlooking App Store restrictions. Changing billions of preferences is nearly impossible; opening a platform’s choke point can directly expand competition and innovation. His analogy was unforgiving: an environmentalist who opposed nuclear power for years has not seriously confronted the relevant trade-offs.
Sharp pushed back that critics of aggregators generally oppose platform abuse too. Thompson’s answer—“The priority order matters”—was that professing concern for everything avoids the revealing question of where advocates actually spent their time and political capital.
4. Apple can charge for its checkout, but not silence commerce
Sharp argued that anti-steering rules and the continuing 30% in-app commission suppress better economic outcomes by taxing digital businesses with no practical mobile alternative. Spotify supplies the subscription’s value, in his framing, while Apple collects a perpetual share merely because the customer owns an iPhone.
Thompson separated two issues Sharp had joined. Apple created a seamless in-app purchase experience and may charge developers who choose it; developers have no entitlement to Apple’s integrated checkout when they can transact on their own websites. What “disgusts everyone” is Apple preventing developers from telling users that an external purchase option exists.
The commission itself has no discoverable neutral rate. Apple’s control over essential users supports a theoretical 99.9% charge, yet if commissions were prohibited, the market-clearing price would be zero because Apple would still need APIs, documentation, and apps to sell iPhones. “The correct rate is either 99.9 or zero. Any point in between is arbitrary.”
Thompson called the judge “100% morally right” but considered parts of the economic and legal reasoning untethered from a principled valuation. His preferred rule is narrower: a mass-consumer platform may control its own device and checkout, while enterprise contexts differ because buyers are more sophisticated and alternatives are more viable. A platform cannot police developers’ speech or unrelated web commerce. He gave Apple roughly a 50/50—“maybe 55/45”—chance on its appeal and stay request.
5. Apple’s profit optimization has eroded its resilience
Apple receives roughly “$20 to $25 billion or whatever the number is” in pure-profit Google search payments, giving Eddy Cue a clear incentive during Google’s remedies hearing. Apple must argue that search is newly competitive while preserving the partnership supporting what Thompson described as perhaps a quarter of its profit.
Thompson connected that dependence to China exposure and App Store intransigence: Tim Cook has optimized every aspect of Apple’s business so aggressively that the company is more susceptible to disruption than an integrated device company should be. Using the oak-and-reed parable, his conclusion was stark: “Tim Cook has destroyed Apple’s resilience.”
Apple consequently has little voluntary incentive to make AI search the primary Safari experience. Thompson expects it to deliver the option that monetizes best, not necessarily the best product, while effectively pleading, “Please Google, hold on.” Sharp’s counterpoint is that this conflict makes judicial intervention unusually valuable.
6. Falling Safari searches turn AI disruption into measured behavior
Sharp relayed Cue’s testimony that April Safari searches fell for the first time in two decades and that OpenAI, Perplexity, and Anthropic might eventually replace conventional engines. Thompson viewed the decline as the proof point markets had awaited: active search behavior is falling even though people still use their phones every day.
The substitution therefore cannot be explained by elongated iPhone replacement cycles or used-device adoption. Users are doing something else on the same phones, and AI is “a one-way ratchet”: unlike Instagram, it is not an entertainment habit people eventually recognize as wasted time and abandon.
Sharp called the possible end of Google’s default placement an “asteroid” moving closer to Search. Thompson initially judged Apple more exposed because losing a quarter of profit is immediate, whereas historically users could simply reselect Google; the new uncertainty is whether an AI alternative could become “utterly better.”
Their disagreement centered on coercion versus market opening. Sharp saw ending default payments as textbook antitrust arriving when credible alternatives finally exist. Thompson said users can already change browsers, choose search engines, or install ChatGPT, and accused Sharp—partly antagonistically—of being “mad that people don’t do what you want them to do.”
7. Google’s ad machine could outlast a weaker search product
Thompson would not make a clean bearish call on Google. It retains commercial queries, an advertising stack that is hard to reproduce, rising CPMs, and a “flight to safety” advantage; clicking ads may also preserve commercial intent that chat interfaces have not yet replicated convincingly.
Thompson called AI Overviews the world’s most-used ad product and noted that more people use generative AI on Google than on any other platform. Google says ads are present and monetization matches Search, leaving a credible path to adapt: incumbencies usually last longer than expected, and new layers often supplement rather than erase them.
The product threat remains obvious: “Search on Google sucks, compared to ChatGPT, for a lot of things.” Yet ChatGPT can resemble Wirecutter—well-presented and sufficiently convenient without necessarily producing the individually optimal recommendation—so interface superiority does not eliminate questions about personalization or commercial discovery.
Thompson’s $2,000 microphones supplied the concrete example. Generic reviewers overweight price, while professional audio is cheap relative to his livelihood; an audio professional identified the required headset and interface. He wants AI to infer such priorities from purchase history, making privacy “a massive detriment to the user experience” when it prevents a system from knowing him.
8. ChatGPT exposes college as a credentialing machine
Thompson had framed ChatGPT through “AI Homework” on December 5, 2022, and found the 2025 panic eminently foreseeable. His thesis: technologies like this are “less causal than they are revelatory”—ChatGPT is not making most college worthless so much as revealing how little educational substance many programs required.
He still insists parents should send children to college because the wage premium and credential remain significant. The distortion is extending society’s qualifying checkpoint from high-school graduation around age 18 to college at 22, then graduate credentials at 24 or 25, with debt, fertility, and other downstream consequences.
Thompson argued that employers, constrained from directly testing applicants after decisions such as Griggs v. Duke Power, outsourced filtering for conscientiousness and baseline intelligence to universities. Policymakers then reversed cause and effect: because people passing the filter earned more, society assumed pushing more people through degrees would itself create higher earnings.
9. Real learning survives AI, but assessment must become honest
Thompson’s own political-science degree was easy, yet self-organized five-person seminars reading Plato and debating education made his final two years extraordinary. For intrinsically motivated students, AI extends that experience: he opens numerous chats because every unknown fact or argument triggers a desire to investigate further.
His perfect-competition analogy was sportswriting. Every city once supported a columnist because readers were geographically captive; once the internet removed that constraint, audiences discovered perhaps only ten were nationally worth reading and flocked to figures such as Bill Simmons. AI similarly exposes quality differences institutions once concealed.
Sharp preserved the strongest objection: moving from a blank page to several semi-coherent pages teaches effort, and students who invoke ChatGPT at the first difficulty may never develop critical thought. The article’s brutal specimen was a student spending roughly ten hours a day on TikTok, until her eyes hurt, while AI reduced an essay to two hours.
Their practical answer was old-fashioned assessment: supervised written exams and blue books remain available. Thompson would also ban classroom laptops because handwritten notes improve retention. He conceded that replacing honor-based take-home work with hard controls is a genuine loss—when shared mores disappear, rules and red tape must take their place.
10. AI rewards agency while institutions punish honest effort
From an employer’s perspective, knowing how to use AI may be exactly the skill college should certify. Thompson characterized much coursework as busywork and noted that the student building cheating applications displayed entrepreneurial initiative, even though large employers usually want “cogs in the machine.”
His Microsoft experience made that conflict personal. Rebuilding a superior deck earned worldwide adoption but no promotion because it bypassed hierarchy; the impatience and independence that made him a poor corporate cog later enabled Stratechery and a podcast with tens of thousands of listeners.
AI could give more high-agency individuals the leverage corporations once reserved for founders. “Apple was a Formula 1 car built around Steve Jobs. NVIDIA is a Formula 1 car built around Jensen Huang”; AI might supply comparable amplification without tens of thousands of employees. Aggregate GDP could rise even while the distributional experience resembles trade’s gains and Rust Belt losses.
The darker endpoint is a majority with food, housing, and endless entertainment but little purpose, while a smaller group compounds initiative into enormous output. Thompson called them people “self-consigning themselves to a world of being cattle,” yet Sharp stressed the institutional failure too: honest students graded on a curve are now “competing against the AI” and being punished for doing the work.