Apple is taking a $1 billion-a-year gamble, betting on Google’s Gemini AI to fix Siri while simultaneously building a “walled-off” garden to protect its data. This “interim solution” will see the 1.2 trillion parameter model integrated into the new “Linwood” Siri.
This move, part of the “Glenwood” project, is a tacit admission that Apple’s 150-billion parameter models are not enough. Google’s “ultrapowerful” AI, which won a “bake-off” against OpenAI and Anthropic, will now handle all complex “summariser” and “planner” functions.
The hybrid system, launching in the spring, will use Apple’s tech for simple tasks and Google’s for complex ones. This is a reluctant move for Apple but a necessary one to stay competitive.
The deal’s non-negotiable aspect is privacy. The Gemini model will run on Apple’s Private Cloud Compute servers, not Google’s. This architecture ensures Google gets its $1 billion check but is completely locked out of Apple’s user data ecosystem.
While Apple’s management pushes its teams to build a 1T+ model replacement, this “behind-the-scenes” partnership, overseen by Craig Federighi and Mike Rockwell, could define the two companies’ AI relationship for years.
