Apple has no ChatGPT of its own. No frontier model. Siri still lags behind the competition. Vision Pro flopped. The AI chief quit. The CEO is leaving in September. And yet they just closed a record quarter — $111.2B in revenue (+17%), $29.6B net income (+22%), market cap near $4 trillion. Stock up 3% after the report. Something doesn't add up, I thought.
Oh, and Tim Cook is out, handing the company to John Ternus — an engineer who's spent 25 years at Apple (he's got a fun LinkedIn profile, literally one line on his résumé — Apple — and he only set his profile picture a couple days ago). So here's the thing: Apple is at its peak and supply-constrained at the same time. Mac mini, Mac Studio, and MacBook Neo have been backordered for months. On the analyst call, Cook said "both of these platforms are very well suited for AI and agentic tasks, and customers understood that faster than we projected." They're bottlenecked by TSMC capacity on advanced nodes — the same lines that print chips for AI servers. Memory is getting more expensive. Cook warned of "meaningfully higher memory costs" this quarter and beyond.
Meanwhile, 2026 capex on AI: Microsoft and Google, ~$190B each. Meta, $125–145B. Amazon, $200B. Apple, $14B. Analysts read this gap not as falling behind, but as a deliberate bet. Apple isn't building its own frontier AI. Apple owns 2.5 billion devices through which someone else's AI reaches people. The frontier model? Licensed from Google: Gemini, roughly $1B a year, about 1% of free cash flow.
That billion is already paying for itself. Per AppMagic, in 2025 Apple collected $900M in App Store commissions from generative AI apps alone. In 2026, the forecast is over a billion. ChatGPT drove 75% of that through in-app subscriptions. OpenAI, Google, Anthropic, and xAI build the AI. Apple takes 30% of every subscription, because iPhone is the primary delivery channel to the end user.
Analysts are expecting another layer this fall: Apple Intelligence as a paid subscription. At 2.5 billion devices, even modest conversion rates mean billions in recurring revenue stacked on top of a $31B services business per quarter.
In parallel, Apple is working on AI glasses — a competitor to Meta Ray-Ban. Two cameras, hand gesture controls, integration with an updated Siri, no screen in the first version. A preview is possible by year-end, launch in 2027. Vision Pro, meanwhile, is effectively frozen: the M5 chip refresh generated no excitement, and the $3,500 headset never became a mass-market product.
Apple has plenty of problems. Siri still can't do what ChatGPT does. No homegrown AI breakthrough in sight. The hardware shortage will pressure revenue for a few more quarters. But the company has done something none of the hyperscalers have managed: it made money from the AI boom without going into debt or burning hundreds of billions on infrastructure. It made money on its core products — iPhone, Mac, App Store. A $4 trillion market cap — no proprietary frontier model, no data centers stacked with hundreds of thousands of GPUs, no splashy AI demos. Probably the most boring and most intelligent bet in this entire race.