Apple’s ‘Private Cloud Compute’ to Host Google’s 1.2T Gemini AI

The centerpiece of the $1 billion-a-year deal between Apple and Google is Apple’s “Private Cloud Compute.” This “walled-off” architecture is what allows Apple to use Google’s 1.2 trillion parameter Gemini AI for Siri without compromising its privacy-first stance.
The Gemini model will be hosted on Apple’s servers, not Google’s, ensuring Google never sees Apple user data. This “interim solution” was necessary for Apple’s “Glenwood” project, the internal effort to fix Siri.
Google’s “ultrapowerful” AI won a “bake-off” against OpenAI and Anthropic, proving it was the only model capable of the “summariser” and “planner” functions Apple needs for its new “Linwood” assistant.
This hybrid system will use Apple’s 150-billion parameter models for simple tasks and Google’s AI for complex ones. It’s a reluctant admission of Apple’s AI lag, overseen by executives Craig Federighi and Mike Rockwell.
While Apple pushes its teams to build a 1T+ replacement, this “temporary” fix could last for years, a testament to Google’s AI lead and Apple’s creative privacy solution.