Apple’s Siri Overhaul Relies on Google TPUs and Nvidia Chips
AAPL•At WWDC, Apple revealed that key AI models for its upgraded Siri were distilled from Google’s Gemini and partly trained on Google’s TPUs, with high-end AFM Cloud Pro running on Google Cloud using Nvidia chips. This deepens Apple’s reliance on Alphabet, potentially increasing Google’s negotiating power and raising regulatory scrutiny.
1. WWDC AI Collaboration
At WWDC, Apple disclosed that its advanced Siri and AI-enhanced apps leverage models distilled from Google’s Gemini, with portions of training executed on Google TPUs to accelerate development and performance.
2. AI Infrastructure Deployment
Apple’s high-end AFM Cloud Pro model will operate on Google Cloud infrastructure using Nvidia chips for compute-intensive tasks, while simpler requests will process locally on iPhone devices to optimize responsiveness and user privacy.
3. Negotiating Power and Regulatory Concerns
This deeper integration could amplify Google’s leverage over Apple, supplementing the existing $20 billion search engine default agreement and potentially inviting regulatory scrutiny over privacy, data handling, and competitive dependencies.
