Apple's AI privacy strategy now depends on Google and Nvidia infrastructure, complicating its core pitch

Apple Intelligence features and updated Siri AI are designed to process queries on-device where possible, and otherwise through Apple's Private Cloud Compute system, with Apple claiming data won't be stored, used beyond request execution, or accessible to Apple or third parties.

Private Cloud Compute has expanded from running on Apple silicon in Apple's own data centers to also running on Google Cloud systems using Nvidia GPUs, Intel CPUs, and Google Titan chips—a shift from Apple's original emphasis on a hardened supply chain with extensive security validation before each server joins the rack.

Apple collects only 'limited information' on Private Cloud Compute requests such as size and completion time, but no content or results, and says it doesn't use private data or user interactions to train its foundation models—a substantially narrower collection than competitors: Google Gemini collects prompts, files, recordings, and content by default with 18-month chat history retention; OpenAI collects prompts and device information with chats used for training by default; Anthropic collects similar data with retention up to five years.