Models & technology

Private Cloud Compute

People & organizations
Apple

Private Cloud Compute runs AI requests too heavy for the device on Apple silicon servers, designed so that personal data is not retained or accessible to Apple.

Cloud processing built for privacy

Private Cloud Compute (PCC) handles AI requests that are difficult to process on the device, running them on dedicated Apple silicon servers. It extends Apple Intelligence into the cloud while aiming to keep privacy protections close to what iPhone and Mac provide locally.

The design does not retain the personal data sent to it, and Apple states it is not accessible to Apple. There is also a mechanism for independent experts to verify the software running on the servers.

When it appeared, and what changed

Private Cloud Compute was announced at WWDC24 alongside Apple Intelligence. It is not new in WWDC26.

WWDC26 kept it as the place the next generation of Apple Foundation Models runs, and added a way for developers to reach PCC models through the Foundation Models framework.

How it differs from ordinary cloud AI

Cloud AI services encrypt traffic and manage data too, but the provider runs the servers and may retain logs and input for a period. PCC emphasizes sending only the data needed for the request and being able to verify technically that it is not kept afterward.

This does not mean avoiding the cloud. The request travels from the device to Apple’s servers, which distinguishes it from fully on-device AI.

Worth checking

Which requests stay on the device and which go to PCC depends on the feature and the model. When processing is handed to an outside AI service, different privacy terms apply.

For sensitive material, confirm before sending whether Apple or a third party will process it.

Related terms

Sources and review information

Last reviewed July 17, 2026

Back to the AI Glossary

Search this site