When Bloomberg reported in summer 2025 that Apple was considering replacing its in-house AI models with external providers, leadership reacted within hours. Siri chief Mike Rockwell and AI head John Giannandrea convened an emergency meeting with the Foundation Models team. Rockwell’s message: the article was “bullshit.”

Apparently, no one bought it. Apple Intelligence was lagging competitors, and the revamped Siri was delayed. Over the following months, talent began leaving—among them team lead Ruoming Pang—many of them recruited by Meta.

Billion-dollar demands and strategic conflicts

Apple held parallel talks with Anthropic and OpenAI, and both paths fell apart. Anthropic reportedly demanded several billion dollars per year over multiple years. OpenAI posed a different problem: the company was actively poaching Apple engineers and, with ex-Apple designer Jony Ive, pursuing its own hardware ambitions.

Google wasn’t initially the favorite, but ultimately won out. It wasn’t viewed as the clear tech leader, and an antitrust case threatened the Safari search deal. In September 2025, however, a judge ruled the deal could stand. Google also offered terms Apple found acceptable.

A year-long power struggle ended with Giannandrea’s exit

Behind the scenes, a power shift had already begun in early 2025. CEO Tim Cook reportedly lost confidence in Giannandrea. Software chief Craig Federighi effectively took control of Apple’s AI direction and pushed for external partners.

Giannandrea was internally sidelined for nearly a year before Apple publicly announced his departure in December 2025. He will keep his salary and equity through April 2027.

His exit affects several initiatives: an AI-powered Safari effort is partly on ice, and a ChatGPT-style project called “World Knowledge Answers” has been sharply scaled back. In its place, Gemini is now set to replace internally developed technology—models Apple reportedly refers to internally simply as “Apple Foundation Models.”