What Is Private Cloud Compute? Apple AI Requests Without Retaining Your Data

Most cloud AI keeps a copy of whatever you send it, somewhere, for some length of time. Apple built Private Cloud Compute to break that assumption. When an Apple Intelligence request is too big to run on your iPhone or Mac, it goes to Apple silicon servers that process the prompt, return the answer, and then hold nothing. Not even Apple staff with admin access to the hardware can reach into what was sent.

Quick Answer

Private Cloud Compute is Apple's server tier for the heavier Apple Intelligence requests that an on-device model cannot handle alone. Your data is encrypted to specific server nodes, used only to produce the response, and deleted the moment the answer is returned, with nothing retained afterward. The software running on those servers is independently auditable, so security researchers can verify the privacy promise rather than take it on trust.

How a request actually flows

Apple Intelligence starts on your device. A local model decides whether your request can be answered right there, and many small tasks never leave the phone at all. When a request needs a larger model, your device builds a package containing just the prompt and the parameters needed to run it, then encrypts that package directly to the public keys of the specific Private Cloud Compute nodes it has confirmed are genuine and cryptographically certified. The data is not sent to a general Apple cloud and then sorted. It is aimed at verified hardware before it ever leaves your hands.

Once the node finishes, it sends the response back and the data is gone. There is no logging of your prompt to a database, no copy kept for training, and no admin console where an engineer could pull up what you asked.

What makes the servers different

The foundation is Apple silicon, the same chip family that runs the Mac and iPhone, carried into the data centre. That brings the Secure Enclave to the server to protect the encryption keys, and Secure Boot to guarantee the operating system loaded onto each node is signed and verified exactly as it is on iOS. A server that cannot prove it is running the correct, unmodified software is not trusted to receive your encrypted request in the first place.

The piece that sets it apart from rival cloud AI is verifiability. Apple publishes the software images that run on these nodes so that independent privacy and security researchers can inspect the code at any time. The claim is not just that Apple chooses not to keep your data, but that the system is built so Apple cannot keep it, and outsiders can check the work. If you are weighing up Apple silicon machines, the current MacBook lineup at Evetech is the same chip architecture that anchors this whole approach.

Does it matter for everyday use

For most people the benefit is quiet. You get the answer quality of a large cloud model without handing a permanent copy of your request to a server farm. That is a meaningful gap from the typical assistant, where prompts may be stored and reviewed. If privacy is a deciding factor in your next machine, it is worth comparing what sits near the top of the best-selling laptops at Evetech against an Apple silicon option that ties into this model.

Frequently Asked Questions

Does Apple keep my Apple Intelligence requests?

No. Requests sent to Private Cloud Compute are used only to generate the response and deleted once it is returned, with no copy retained in any form afterward. On-device requests never leave your device at all.

Can Apple staff read what I send to Private Cloud Compute?

No. The system is designed so user data is never available to Apple, including staff with administrative access to the production servers or the hardware itself.

How is this different from other cloud AI assistants?

Typical cloud assistants may store prompts, sometimes to improve their models. Private Cloud Compute deletes data after each request and publishes its server software so independent researchers can verify that the privacy guarantees hold.

Do I need a specific device to use it?

Private Cloud Compute works behind the scenes on Apple devices that support Apple Intelligence. You do not switch it on manually. The device decides when a request needs the server tier and handles the encryption automatically.