NVIDIA Confidential Computing powers Apple's Private Cloud Compute, expands to Google Cloud

NVIDIA's Confidential Computing GPUs now underpin confidential server-side inference in Apple's Private Cloud Compute (PCC), the privacy-preserving cloud layer Apple uses when on-device models aren't enough. The notable expansion: PCC is moving beyond Apple's own data centers onto Google Cloud, reflecting the Apple–Google model partnership announced at WWDC where Apple's third-gen Foundation Models are built with Gemini.

The technical hook is Confidential Computing — hardware-isolated execution environments on the GPU that keep data and model weights encrypted and inaccessible even to the cloud operator. For Apple, this is essential to preserving its privacy promise while running larger models server-side; it lets Apple claim that even when queries leave the device, no one (not Apple, not Google) can inspect the data.

Strategically, this cements NVIDIA as the trusted-execution layer for privacy-sensitive AI inference, an increasingly important differentiator as regulation and enterprise security demands grow. It also deepens the unusual Apple-Google-NVIDIA triangle now powering Siri AI.

The caveat is that confidential computing's guarantees depend on correct implementation and the integrity of NVIDIA's attestation chain — security researchers will scrutinize the actual isolation properties. The HN debate over Apple outsourcing intelligence to Gemini also extends here: routing Apple users' private inference through Google Cloud, even encrypted, raises trust questions Apple will need to address transparently. Readers should watch for technical write-ups on PCC's attestation and any independent audits.