Apple quietly turned generative AI into a serious moneymaker last year, even while looking like it was lagging behind in the AI race. In 2025 alone, AI apps on the App Store paid Apple nearly $900 million in fees – and the wild part is that most of that money came from other companies’ AI models, not Apple’s own.

If you had to point to one app that effectively wrote Apple a massive check last year, it would be ChatGPT. OpenAI’s app alone is estimated to have generated roughly three‑quarters of Apple’s generative AI commission revenue in 2025, thanks to downloads that turned into in‑app subscriptions on the iPhone and iPad. Elon Musk’s Grok was a very distant second, accounting for around 5% of that total, with the rest spread across a long tail of other AI tools – image generators, code assistants, writing helpers, and more.

Behind those big headline numbers is a simple dynamic: if you want mainstream users, you still want to be on the iPhone. AppMagic’s data, cited by multiple outlets, shows that revenue from generative AI apps on the App Store almost tripled between January and August 2025, peaking at about $101 million in a single month before cooling slightly as ChatGPT downloads slowed. Apple’s cut comes from the familiar App Store model — up to 30% commission on digital purchases and subscriptions — so as AI subscriptions exploded, Apple’s services line quietly got a very healthy bump.

That’s especially striking when you compare it to how other tech giants are approaching AI. Microsoft, Amazon, Meta, and Google are spending tens of billions of dollars building out massive data centers and GPU farms to train and run frontier models, often without clear near‑term profit to show for it. Apple, by contrast, has kept capital expenditures relatively flat and leaned into something it already does extremely well: sell hardware at scale and take a platform fee from whatever software runs on top of it. In other words, while rivals fight expensive AI model wars in the cloud, Apple is happily running the toll booth at the device level.

That’s the core of Apple’s current AI strategy: it doesn’t need to own every model if it owns the rails. iPhones remain one of the primary ways people access tools like ChatGPT, Gemini, Anthropic’s Claude, and xAI’s Grok, and every time a user turns a free download into a paid subscription, Apple takes its slice. Some investors have started explicitly describing this as Apple acting like a “toll road” for AI providers – the company lets others build the cars and engines, then charges them to drive through its ecosystem. For Apple, that’s an attractive position: lower technical risk, steady high‑margin services revenue, and very little need to chase every new model release.

At the same time, Apple isn’t ignoring AI altogether; it’s just picking its spots. On the consumer side, the company has been pushing an “on‑device first” strategy for Apple Intelligence, using custom silicon and private user data stored on the iPhone to power features like smarter text suggestions, summaries, and image generation without sending everything to the cloud. That approach is cheaper from an infrastructure perspective and plays nicely with Apple’s long‑running privacy narrative, but it also means Apple’s own models are smaller and less powerful than the giant cloud models its competitors are training.

Nowhere is that trade‑off more obvious than Siri. By most modern standards, Siri has felt behind the curve compared to tools like ChatGPT, Gemini, and Claude, especially when it comes to open‑ended reasoning and conversational flexibility. Rather than trying to catch up alone, Apple has effectively decided to rent someone else’s brain: it struck a deal with Google to use a custom 1.2 trillion‑parameter Gemini model to power a revamped Siri and future Apple Intelligence features. Reports put that deal at roughly $1 billion per year, giving Apple access to a model about eight times larger than its own current cloud‑based systems.

There’s a nice bit of irony baked into that relationship. Google already pays Apple an estimated $20 billion annually to remain the default search engine on iPhones, a deal that has been a cornerstone of Apple’s services revenue for years. Now money is starting to flow in the other direction as Apple pays Google for cutting‑edge AI infrastructure, albeit at a much smaller scale. When you zoom out, you get an almost circular economy: Google funds Apple’s search default, third‑party AI apps fund Apple via App Store fees, and Apple uses part of that services windfall to pay Google for Gemini so Siri can become competitive again.

Regulators are increasingly poking at this model, and that’s a subplot worth watching if you care about where this all goes next. Apple has already started to tweak App Store commissions in some regions under pressure — for example, lowering fees in China from 30% to 25% for many developers, and reducing rates for smaller devs and certain subscriptions. In Europe and the US, ongoing legal and regulatory scrutiny has raised questions about whether the classic “Apple tax” can stay as high as it is today, especially for digital content and subscription apps. If those pressures intensify, the economics of Apple’s AI toll road could change, even as user demand keeps growing.

Still, nearly $900 million from generative AI apps in a single year – with the trajectory pointing toward $1 billion and beyond – shows how powerful Apple’s position as a platform really is. Even without leading the pack on frontier models, it has managed to become one of the biggest beneficiaries of the AI boom simply by controlling the devices and the store where most people first experience these tools. As Google, OpenAI, Anthropic, xAI, and others battle over who has the smartest model, Apple has quietly built the place where those models meet hundreds of millions of paying users – and for now, that might be the most profitable AI strategy of all.