Apple turns the operating system into AI's distribution layer

Apple Intelligence puts free on-device models and ChatGPT inside iOS. Distribution now belongs to the platform, and AI apps must offer more than model access.
Generative AI is now a free part of the operating system on the largest premium device platform. Apple introduced Apple Intelligence on 10 June, a set of generative models built into iOS 18, iPadOS 18 and the next release of macOS, and its shares rose about 7% the next day to a record close.
Giving AI away as a feature, rather than selling it as a product people pay for separately, shifts bargaining power to whoever owns the device. Model makers now compete for a slot inside the platform, and Apple can add or swap providers. Value in consumer AI concentrates in the companies that own the device and the personal context, while the price of generic features falls toward zero.
The architecture shows how much Apple intends to control. Many requests run on the device itself. Harder ones go to Private Cloud Compute, Apple's own servers on Apple silicon, which Apple says retain no user data and whose code independent experts can inspect. Running that tier on its own chips is a sign that platforms want control over inference cost and user data. Siri and the system writing tools can also pass a request to ChatGPT, powered by GPT-4o, free and without an account, and Apple plans to add other model providers. The features arrive in beta this fall in US English and need an iPhone 15 Pro or a Mac or iPad with an M1 chip or later.
Writing help, image generation and summaries of mail and notifications, categories that independent apps charged for, now ship with the phone, the squeeze the GPT Store started. A consumer app whose main value is access to a capable model faces a competitor that is free, built in and aware of the user's messages, calendar and photos. The opening is in depth: workflows inside a profession, data the platform does not hold, and services that run across devices and companies. Apple also lets third-party apps expose actions to Siri, which makes the assistant a new channel, so products for iPhone should be designed for the operating system to call them and priced for outcomes that a general assistant cannot deliver.
Two numbers will size that channel. One is whether the hardware requirement drives an iPhone upgrade cycle this fall, and the other is how fast the features reach languages beyond US English, which sets the size of the market for apps built on top. The providers Apple adds next, and on what commercial terms, will show whether model makers pay for distribution or get paid for capability.
Platform owners are claiming the general assistant. Fanaura's thesis rests on depth rather than model access, on AI-native companies that sit deeper in work and daily life than a general assistant can reach, with data and distribution of their own. The phone now owns the assistant, and what stays open is everything it cannot reach.
Building an AI-native company? Apply in five minutes. →
More from Fanaura
Building an AI-native product that goes deeper than a built-in assistant can? Send us the project.
Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.
Get the news by email
One short note each weekday. No marketing, and you can unsubscribe at any time.