Apple Concedes AI Lag, Pays Google $1B for ‘Interim’ Siri Brain

In a clear sign of its current standing in the AI race, Apple is paying its rival Google $1 billion a year to use its Gemini AI. This “interim solution” will see Google’s 1.2 trillion parameter model become the “behind-the-scenes” powerhouse for complex tasks on the new Siri.

Apple’s internal effort to fix Siri, project “Glenwood,” identified a massive capabilities gap. Apple’s own 150-billion parameter models couldn’t compete with Google’s “ultrapowerful” AI, which won an extensive “bake-off” against other top models from OpenAI and Anthropic.

The new “Linwood” Siri, due in the spring, will be a hybrid. Simple requests will use Apple’s tech, but the advanced “summariser” and “planner” functions needed for multi-step tasks will be handled by Gemini.

This is a reluctant move for Apple, but one pushed by top executives Craig Federighi and Mike Rockwell to make Siri competitive. Meanwhile, it’s a huge win for Google, establishing it as a key “AI supplier.”

To maintain its privacy-first stance, Apple is not letting Google’s AI phone home. The Gemini model will be hosted on Apple’s “walled-off” Private Cloud Compute servers, ensuring Google gets its money but never sees Apple’s user data.