THE MOST LUCRATIVE artificial-intelligence strategy of 2025 may belong to the company with the worst chatbot. Apple is on pace to surpass $1 billion in AI revenue this year — not by building a frontier model, not by deploying a fleet of data centers, but by collecting a commission every time someone subscribes to ChatGPT through an iPhone. It is the kind of elegant arbitrage that would make a Renaissance toll collector blush.
The mechanics are straightforward, if quietly ruthless. Generative AI apps paid Apple nearly $900 million in App Store fees last year, according to AppMagic, the analysis firm. Three-quarters of that haul came from a single source: OpenAI's ChatGPT. Grok, xAI's chatbot, accounted for roughly 5%. The App Store's standard terms — approximately 30% of subscription fees in the first year, dropping to 15% thereafter — apply to AI apps the same way they apply to meditation timers and calorie counters. However sophisticated the underlying model, Apple's cut is the same.
As a share of Apple's total revenue, a billion dollars barely registers. But within the services business — the segment investors have fixated on for its superior growth rate and fatter margins relative to hardware — generative AI apps represent one of the few categories still accelerating. Monthly commissions from GenAI apps climbed from around $35 million in January 2025 to a peak of $101 million in August, before retreating as ChatGPT downloads cooled.
The gatekeeper's gambit
Yet the more interesting story is not what Apple earns today but what its position permits it to defer. While Amazon, Microsoft, Alphabet, Meta, and Oracle have embarked on a capex arms race — quarterly spending by that cohort now routinely tops $70 billion — Apple's capital expenditures remain a fraction of its rivals'. The company is betting that a combination of on-device inference, custom silicon, and access to the deeply personal data people store on their iPhones will eventually produce an AI experience worth having, without requiring the infrastructure binge that has made Wall Street increasingly nervous about hyperscaler returns.
"If they can act as a toll road for providers of AI, then they'll probably end up looking good long-term for not having the big capex overhang," reckons Charles Rinehart, chief investment officer at Johnson Asset Management, an Apple shareholder. It is a tidy thesis — albeit one that depends on Apple eventually delivering something more than tolls.
The trouble is that Apple's own AI efforts remain conspicuously unfinished. Siri, the company's long-suffering voice assistant, still cannot remember a conversation, conduct research, or generate content in the manner of modern chatbots. The departure of Apple's top AI executive, John Giannandrea, last year underscored how persistently the company has struggled to modernize its assistant. Apple and Google announced in January that Gemini would power a revamped Siri later this year — an admission, however diplomatically framed, that Cupertino could not get there alone.
Meanwhile, OpenAI is making its own play for hardware relevance. Its acquisition of a startup co-founded by Jony Ive and other former Apple designers signals an ambition to bypass the App Store entirely — to own the device, the model, and the distribution. Google's Pixel phones already ship with AI capabilities Apple cannot yet match, though those features have not proved compelling enough to trigger meaningful switching from iOS.
The competitive dynamics, then, are unusually symmetrical. OpenAI wants to become Apple (owning the device). Apple wants to become OpenAI (owning the model). Google, characteristically, is attempting both simultaneously. Each faces the other's core competency as its own most formidable obstacle. Building a world-class consumer device ecosystem is hard; building a frontier model is hard; the question is which is harder to replicate under pressure.
For now, Apple's position is enviable precisely because it is patient. The company collects a growing tithe from every competitor's AI ambitions, funds its own research at a fraction of the industry's cost, and retains the one asset no rival can easily reproduce: more than a billion active devices in consumers' hands. Whether Cupertino can convert that distribution advantage into genuine AI capability — before someone else converts AI capability into a distribution alternative — is the billion-dollar question. Possibly, in Apple's case, quite literally. ■
For more, join 75,000 subscribers getting tech's favorite brief here
