At this year’s WWDC developer conference, Apple announced that its Private Cloud Compute (PCC) service has officially broken through the boundaries of its own data centers, expanding for the first time to Google Cloud Platform. As the core infrastructure for this partnership, NVIDIA will provide Blackwell GPUs equipped with Confidential Computing technology to support server-side inference for Apple Foundation Models and AI models jointly customized by Apple and Google. It is reported that this underlying architecture deeply integrates technologies from Google’s Gemini series of models and will be used to power next-generation features in Apple Intelligence. The core of this upgrade lies in NVIDIA’s confidential computing technology. This technology isolates AI workloads within Trusted Execution Environments (TEEs) via hardware-level security architectures and strictly verifies the integrity of the underlying infrastructure using cryptographic methods before sensitive data is uploaded. For users, this means that even when conversations or private data are processed on the cloud, the cloud service provider cannot access user information. NVIDIA stated that this move reflects deeper transformations in AI infrastructure: as AI experiences increasingly integrate edge and cloud computing power, the industry must establish strict privacy and security boundaries while ensuring high-performance inference. The implementation of NVIDIA’s confidential computing relies primarily on four capabilities: a hardware-based root of trust ensures that GPUs have not been tampered with; encrypted communication links protect data transmission between components; remote attestation mechanisms allow software to verify platform safety status before releasing sensitive data; and there is native support for accelerating AI inference and training. In an era where demands for data privacy continue to rise, this technology is becoming a critical foundation for trusted AI cloud services.