Apple's AI strategy is failing on five fronts, per Ben Thompson analysis
May 23, 2025
Key Points
- Apple's AI strategy is constrained by internal skepticism: its software chief didn't believe in AI, its AI head was skeptical of LLMs, and its former CFO refused sufficient GPU funding.
- Apple's privacy-first approach blocks access to its own high-value user data for training, a competitive advantage that Meta, Google, and YouTube possess.
- Ben Thompson argues Apple should monetize AI real estate by opening on-device models to developers, offering cloud inference at cost, and allowing third-party AI replacements for Siri with revenue sharing.
Summary
Apple's AI strategy is failing across five structural dimensions, according to Ben Thompson's analysis. The company's head of software engineering did not believe in AI. Its head of AI was skeptical about LLMs and chatbots. The former CFO refused to commit sufficient GPU funding. Apple's commitment to on-device privacy locked the company out of using its own high-value user data for training, an advantage that Meta, Google, and YouTube all possess. The AI team is understaffed, with talent retention a concern given opportunities elsewhere.
Thompson argues Apple should reverse course on closed-garden thinking. The company should open on-device models to developers as an API without restriction, offer cloud inference at cost, and allow Siri to be replaced entirely by third-party AIs with Apple taking a revenue share, starting at 50%. This mirrors Apple's existing search deal: if the company cannot win the AI race outright, it can at least monetize the real estate.
Apple's historical playbook of owning the entire stack and controlling the user experience has collided with a generative AI moment where being closed off costs more than it saves.