Apple’s $1B Concession: Google Gemini Chosen Over OpenAI for Siri

In a major concession of its AI standing, Apple has agreed to pay Google $1 billion annually. The deal is to use Google’s 1.2 trillion parameter Gemini model for its new Siri, following a “bake-off” where Gemini beat OpenAI’s ChatGPT and Anthropic’s Claude.

This “interim solution” is part of Apple’s “Glenwood” project to fix its struggling assistant. The new “Linwood” Siri will be a hybrid, using Google’s “ultrapowerful” AI for complex “planner” and “summariser” tasks that are beyond Apple’s 150-billion parameter models.

This is a significant win for Google, solidifying its role as an “AI supplier” even to its main rival. For Apple, it’s a reluctant move, overseen by Craig Federighi and Mike Rockwell, to stay in the AI game.

Apple is pushing its own teams to build a 1T+ model replacement, but this “temporary” fix is expected to last for years given Google’s rapid advancements.

Privacy is non-negotiable. The Gemini model will be hosted on Apple’s “walled-off” Private Cloud Compute servers, not Google’s. This ensures Google gets its $1B fee but no access to Apple user data.