Nvidia said its confidential computing graphics processing units are being used for confidential inference in Apple Private Cloud Compute (PCC) as it expands the system to Google Cloud.

The company said the chips will support server-side inference for Apple Foundation Models, which Apple separately mentioned were developed with Google using the technology behind Gemini models.

Apple originally described PCC as a cloud AI system built with custom Apple silicon in its own data centers.

Apple said the system was designed so user data is processed statelessly, is not accessible to Apple, and can be audited through published software images.

Google Cloud already offers confidential virtual machines with H100 graphics processing units and confidential Google Kubernetes Engine (GKE) nodes for graphics processing unit workloads, showing Nvidia’s role in protecting data while it is being processed.

In a separate expansion note, Apple said PCC’s privacy model remains exactly the same on Google Cloud, including no privileged runtime access and verifiable transparency.

Apple added that it keeps control of the PCC software trusted by user devices while publishing binaries and research tools for external review.

🔗 Source: Nvidia

Recent Apple developments