The $1 Billion Price Tag: Apple Bets Big on Google’s AI for Siri’s Future

Apple is reportedly placing a $1 billion annual bet on Google, finalizing a deal of that value to license its “ultrapowerful” 1.2 trillion parameter Gemini AI model. This massive expenditure is aimed at one thing: saving Siri. The partnership is designed to completely rebuild the voice assistant’s core technology, providing an “interim solution” that will power a new slate of advanced features set to launch next spring.
This deal is the culmination of Apple’s “Glenwood” project, an internal initiative to fix the lagging assistant. After testing models from OpenAI and Anthropic, Apple’s leadership, including software chief Craig Federighi, chose Google’s Gemini as the most capable short-term fix. The new Siri, “Linwood,” will use this 1.2 trillion parameter model for its “summariser” and “planner” functions, a leap in complexity from Apple’s current 150-billion parameter models.
The $1 billion price tag highlights the immense value and computational cost of high-end generative AI. This model will enable Siri to understand complex, multi-part commands and plan their execution, finally catching up to competitors. However, this power will be carefully firewalled. The Gemini model will run on Apple’s Private Cloud Compute servers, not Google’s, ensuring all user data remains “walled off” and private.
This partnership will be a silent one. Apple will not publicly promote Google’s involvement, treating it as a technology supplier, much different from the visible Safari search deal. This arrangement allows Apple to leverage Google’s superior AI without publicly admitting its dependency on a chief rival for such a core feature.
Apple’s long-term goal is to end this dependency. Its AI teams are racing to complete a 1 trillion parameter in-house model to replace Gemini, potentially as early as next year. However, with Google’s Gemini 2.5 Pro already leading the pack and continuously improving, Apple is locked in a difficult and expensive race to catch a moving target, all while paying $1 billion a year for the privilege of using its rival’s tech.