Apple just admitted something interesting at WWDC26: its own data centers aren’t enough. The company announced that its Private Cloud Compute service, originally built to run exclusively on Apple silicon inside Apple-controlled facilities, is expanding to Google Cloud infrastructure powered by Nvidia’s Blackwell GPUs.
What’s actually happening under the hood
Private Cloud Compute launched in 2024 as Apple’s answer to a fundamental tension in AI. On-device models are private but limited. Cloud models are powerful but require trusting the server operator. PCC was designed to split the difference, processing complex AI tasks on remote servers while maintaining privacy guarantees typically associated with on-device computation.
The technical stack reads like a security layer cake. Nvidia’s Confidential Computing provides trusted execution environments on Blackwell GPUs. Intel’s TDX (Trust Domain Extensions) handles CPU-level isolation. Google contributes its Titan security chip technology. Together, these components create encrypted pathways that prevent anyone, including the cloud operator, from accessing data during processing.
Apple retains full control over the PCC software layer. Only cryptographically approved binaries get deployed across the infrastructure, meaning no rogue code can sneak onto the servers. External researchers can still verify the security properties independently, a transparency mechanism Apple established with the original PCC launch.
