Apple AI cloud: Custom-built server hardware based on Apple Silicon, a new OS and Swift-based ML stack

Apple Intelligence relies on both on-device processing and server-based processing. For complex requests that require more processing power, Apple Intelligences takes advantage of Private Cloud Compute, Apple’s own AI cloud. Instead of depending on other hyperscalers like Microsoft, Amazon and Google, Apple built its own servers powered by Apple silicon to ensure that user data is never retained or exposed.

Apple yesterday published a blog post explaining the technology behind Private Cloud Compute. Here are the highlights:

• Apple is using custom-built server hardware that brings the power and security of Apple silicon to the data center.
• Apple developed a new server OS (operating system) based on the foundations of iOS and macOS. The new OS is optimized for Large Language Model (LLM) inference workloads while presenting an extremely narrow attack surface.
• Apple’s AI server stack excluded components that are traditionally critical to data center administration, such as remote shells and system introspection and observability tools.
• Apple created new components that can only provide a small, restricted set of operational metrics to SRE staff.
• Apple used Swift on Server to build a new Machine Learning stack specifically for hosting our cloud-based foundation model.

Apple also mentioned that independent experts can inspect the code that runs on Apple silicon servers to verify privacy. Apple will also assist security researchers by providing the following:

• A Private Cloud Compute Virtual Research Environment: a set of tools and images that simulate a PCC node on a Mac with Apple silicon, and that can boot a version of Private Cloud Compute software minimally modified for successful virtualization.
• Binary images of every production Private Cloud Compute build. Apple will periodically also publish a subset of the security-critical PCC source code.
• Private Cloud Compute images will include the sepOS firmware and the iBoot bootloader in plaintext, making it easier than ever for researchers to study these critical components.

You can learn more about Apple’s Private Cloud Compute here.