Apple is revamping Siri to use Nvidia’s Blackwell B200 processors in Google Cloud. The shift follows assurances that Apple Intelligence would run only on Apple Silicon and Apple‑controlled servers. The partnership, first reported by The Information, also aims to handle more complex language models without compromising response times, while Apple continues to tout its privacy narrative.

Apple’s Private Cloud Compute will encrypt prompts as they traverse the Nvidia chips, a feature dubbed confidential compute. This means only the user’s query leaves the device, and the response is generated without retaining data for training. Apple’s rule forbids storing AI prompts, keeping output private. The encrypted pipeline runs on the chip’s dedicated secure enclave, preventing even cloud administrators from viewing raw input.

Industry observers note the move signals Apple’s willingness to blend its privacy‑first stance with third‑party cloud horsepower. While Nvidia’s hardware promises lower latency, reliance on Google’s servers introduces a shared‑responsibility model for data protection. Apple will showcase the upgraded Siri at WWDC on June 8, confirming the hybrid architecture is now live. Developers will receive SDK updates to integrate the new backend seamlessly.