Apple’s ongoing partnership with Google to use Gemini AI in its data centers is proving more comprehensive than initially disclosed. New reporting reveals Apple has “complete access” to Gemini for training smaller, specialized AI models??otentially a key strategic move as the company races to catch up in the AI race.
The arrangement, announced as part of Apple’s January AI personalization strategy, allows Apple to leverage Google’s massive Gemini model through a distillation process. Google trains “teacher” models; Apple distills that knowledge into smaller, more efficient “student” models optimized for its own devices.
What Distillation Means in Practice
Model distillation is a technique where a smaller neural network learns to mimic a larger one. The smaller model captures the “knowledge” of the larger model in a compressed form that runs faster and requires less compute.
For Apple, this approach offers several advantages: the company gains access to state-of-the-art AI capabilities without building them from scratch, while still delivering something optimized for iPhones, iPads, and Macs. On-device AI processing has privacy benefits Apple has emphasized??ata stays on the device rather than being sent to cloud servers.
The question is whether distilled “student” models can match the quality of their “teacher” models. Research suggests well-executed distillation can preserve 95-99% of a larger model’s capabilities in a much smaller package, but some capability loss is inevitable.
Apple’s AI Strategy: Partnership Over Building
Unlike competitors Microsoft, Google, and Amazon??hich have invested billions in AI infrastructure and training??pple has taken a more pragmatic path. Rather than building massive AI training operations, Apple has focused on integration, user experience, and privacy.
This partnership approach has trade-offs. Apple gets access to cutting-edge AI quickly and cheaply, but it remains dependent on Google’s technology. If Google’s partnership terms change, or if Gemini falls behind competing models, Apple’s AI ambitions could be affected.
Meanwhile, Apple’s competitors continue investing heavily. Microsoft has OpenAI, Google has DeepMind and Gemini, Meta has Llama. Apple, Samsung, and others using Android have watched as iPhone AI features fell behind??nd the Gemini deal is explicitly an attempt to close that gap.
What This Means for Siri
The most visible beneficiary of the Gemini arrangement should be Siri. Apple’s voice assistant has lagged behind Google Assistant and ChatGPT in capability for years. With Gemini-powered distillation, Apple can potentially give Siri much more sophisticated language understanding and reasoning without requiring iPhone-level compute.
Privacy advocates may appreciate that Apple’s distilled models run locally. But there’s an inherent tension: distillation requires access to Gemini’s full capabilities in Google’s cloud. How much user data flows to Google as part of this arrangement isn’t entirely clear from public disclosures.
For iPhone users, the practical result should be a meaningfully smarter Siri capable of handling more complex queries, better contextual understanding, and more useful responses. Whether that’s enough to satisfy users who have watched Apple’s AI fall behind for years remains to be seen.
The Gemini deal represents pragmatic realism from Apple??cknowledge you’re behind, partner with the leader, focus on execution. It’s a strategy that has worked for Apple before (using Intel chips, then switching to ARM; using Google Maps before Apple Maps was ready). Whether it works in AI remains the billion-dollar question.