Apple to Revamp Siri via Google Cloud and Nvidia Blackwell B200 with Confidential Compute
A series of reports from The Information has revealed the technical architecture behind Apple's upcoming AI initiatives for Siri. Apple plans to route a portion of complex Siri queries to Google Cloud, running on a licensed version of Google's Gemini models.
To power these workloads, Apple will tap into Google's fleet of Nvidia Blackwell B200 data center GPUs. Crucially, to uphold Apple's strict privacy standards, the company will enable Nvidia's hardware-based confidential compute feature. This security system encrypts data as it is actively processed on the Blackwell GPUs, preventing even the cloud host (Google) or third parties from accessing user queries or model states in plaintext.
This move marks a major departure from Apple's long-standing strategy of maintaining end-to-end control over all critical hardware and software ingredients. Rather than relying solely on its own custom Apple Silicon servers via Private Cloud Compute (PCC), Apple is embracing third-party cloud infrastructure (Google Cloud) and third-party silicon (Nvidia Blackwell) to support the next-generation Siri.