Investment Review №354. AI lends a hand to the bulls

On July 30, Tim Cook held his final earnings call as Apple’s CEO, and on September 9 made his last appearance at a product launch. He leaves behind a business of an entirely different scale from the one he inherited from founder Steve Jobs in 2011. In fiscal year 2025, the company generated $416bn in revenue, $112bn in net income, and nearly $99bn in free cash flow. Services alone now bring in roughly $109bn a year—over a quarter of total revenue—with paid subscriptions surpassing 1.5bn. Under Cook's tenure, Apple's market cap grew from under $350bn to roughly $4.6tn.

That system now passes to John Ternus, an engineer who joined the company back in 2001. For investors, the key question is not whether the new CEO can deliver another product on the scale of the iPhone—his long tenure in hardware gives him relevant credentials—but whether Apple can extract more revenue per user from its installed base of over 2.5bn active devices. This is a monetization story, not a story about how much the company spends building AI infrastructure—it does not need to win the technology race against OpenAI, Google, or Anthropic.

Apple only needs to remain the primary intermediary between the user and artificial intelligence. The model can belong to someone else; the device, the operating system, the payments infrastructure, cloud storage, the apps, and the user's data all stay inside the brand.

As artificial intelligence becomes embedded in email, photos, calendar, messaging, and workplace apps, the cost of switching platforms rises. That can reinforce customer retention, boost demand for iCloud+ and other services, and simultaneously drive device upgrades. Users may not care which underlying model powers a given query but accessing new AI features will increasingly require a current-generation iPhone or Mac with sufficient on-device compute. Given Apple’s scale, even a modest increase in revenue per user across a 2.5bn-device installed base could translate into tens of billions of dollars in incremental revenue. 

Mac demand is already showing early signs of this dynamic. Supply tightened in the spring as interest in local AI compute increased. The company's distinctive chip architecture allows very large models to run on a single machine. OpenAI has reportedly emerged as a major new Mac customer, assembling Apple machines into large clusters akin to data-center infrastructure. The company's products are not a substitute for Nvidia (NVDA) solutions, but amid ongoing chip constraints and the advantages of Apple’s architecture, the surge in demand makes real sense. 
Investors are increasingly viewing the company less as a contender to build the leading AI model and more as the primary distribution layer through which users access it. Apple controls the device, operating system, apps, data and payments, meaning advances in AI can increase the value of its ecosystem while raising switching costs. Crucially, the company does not need to match the capex-heavy data-center buildout of Oracle (ORCL) and other hyperscalers, where elevated infrastructure spending is increasingly weighing on FCF. Whichever model wins can simply be built into Apple's own platform. That leaves new CEO John Ternus free to focus on what the company does best: the hardware for the next technology cycle, and the ecosystem that surrounds it.