Google’s AI division has scored a massive $1 billion-a-year victory after its 1.2 trillion parameter Gemini model crushed the competition in an extensive “bake-off” held by Apple. As a result, Gemini will now become the “behind-the-scenes” engine for the new Siri.
Apple was looking for an “interim solution” to power its “Glenwood” project, the effort to fix its lagging assistant. Google’s “ultrapowerful” model beat offerings from OpenAI and Anthropic, proving it was the best choice to handle the “summariser” and “planner” functions.
This deal is a reluctant admission of weakness from Apple, whose own 150-billion parameter models can’t compete. The new “Linwood” Siri, planned for a spring release, will be a hybrid system, using Google’s AI for heavy lifting.
This move solidifies Google’s role as a premier “AI supplier,” with other major companies like Snap also using its Vertex AI platform. Apple, meanwhile, is pushing its teams to build a 1T+ model to replace Gemini, but faces an immense challenge.
Privacy was a non-negotiable for Apple. The Gemini model will be hosted on Apple’s Private Cloud Compute servers, not Google’s. This “walled-off” architecture, overseen by Craig Federighi and Mike Rockwell, ensures Google never gets access to Apple’s user data.
