Apple is navigating a complex strategic challenge by finalizing a $1 billion-a-year deal to use Google’s Gemini AI for its next-generation Siri, all while promising to keep user data completely separate from Google. This arrangement will see an “ultrapowerful” 1.2 trillion parameter Google model run on Apple’s own Private Cloud Compute servers. This “walled-off” infrastructure is Apple’s solution to a major paradox: how to use a rival’s superior technology to fix its own AI shortcomings without compromising its bedrock principle of user privacy.
The deal stems from Apple’s urgent need to overhaul Siri, an effort internally code-named “Glenwood.” After testing competitors like ChatGPT and Claude, Apple’s leadership, including software chief Craig Federighi, selected Gemini as the best-in-class interim solution. The new Siri, dubbed “Linwood” and set for a spring release, will use this massive AI model to power its most complex features, such as summarization and multi-step task planning, far exceeding the capabilities of Apple’s current 150-billion parameter models.
This hybrid approach means that while Gemini will handle the heavy lifting, some Siri features will still run on Apple’s in-house models. The $1 billion price tag reflects the sheer computational power of the 1.2 trillion parameter model, which is essential for Apple to deliver the sophisticated, context-aware assistant promised to users. Apple has already allocated the necessary AI server hardware within its Private Cloud Compute system to host and power the model independently of Google’s cloud.
This partnership is substantial but will be intentionally kept quiet. Apple will treat Google as a simple technology supplier, not a co-branded partner, avoiding any public promotion of Google’s involvement. This is a starkly different approach from the Safari search agreement, where Google’s brand is visible. This deal is also separate from earlier, failed talks about integrating Gemini as a standalone chatbot, which Apple has instead done with OpenAI’s ChatGPT as an optional feature.
While this $1 billion deal provides a crucial short-term boost, Apple’s long-term goal remains self-sufficiency. The company is actively developing its own 1 trillion parameter cloud-based model, hoping to have it ready for consumer use as early as next year to replace Gemini. However, this is a race against a fast-moving target, as Google continues to enhance Gemini’s capabilities, making it uncertain if Apple’s “temporary fix” can ever be fully phased out.