Apple said it is expanding its Private Cloud Compute system beyond its own data centers and will begin using Google Cloud infrastructure and NVIDIA GPUs for some Apple Intelligence workloads. The move marks the first time Apple says its cloud-based AI privacy framework will operate on third-party data centers.
The company announced the shift alongside a new generation of Apple Foundation Models, which it says were developed with help from technologies behind Google’s Gemini model family. According to Apple, the models are designed to support Apple Intelligence features across both on-device and cloud-based tasks.
Apple originally introduced Private Cloud Compute, or PCC, in 2024 as a way to extend device-level privacy protections to cloud processing when requests are too complex for on-device models. The system was built on Apple silicon and designed to keep cloud inference stateless, tightly controlled, and transparent to outside researchers. Apple says those core requirements are unchanged even as the infrastructure expands.
In the new setup, Apple says it worked with Google and NVIDIA to run more demanding Apple Intelligence workloads on Google Cloud systems using NVIDIA GPUs. The company says these tasks include agentic tool use and more complex reasoning. Apple also says the deployment uses NVIDIA Confidential Computing, Intel CPUs with TDX, and Google’s Titan chip.
Apple emphasized that it is not relying only on confidential computing features to protect data. The company said it treats the full stack, from firmware through operating systems and application code, as part of its trusted computing base. It also said it has put in place protections against supply chain abuse, including a cryptographically verifiable ledger of hardware in the PCC fleet and software attestation tied to multiple independent roots of trust.
The company added that request parsing will happen in a dedicated process, shared inference software will be rotated with a short lifespan, and attested keys will live in a separate confidential virtual machine isolated from outside inputs. Apple says these measures are intended to preserve the privacy and security properties associated with PCC even when the infrastructure runs outside Apple-owned facilities.
Apple said it will keep control over the software that runs in PCC, and that Apple devices will only trust software that is cryptographically approved by the company. The rollout on Google Cloud will begin gradually during a summer preview period, with the full protections arriving over time.
The company also said the move will not change its transparency approach. It plans to publish all binaries for public inspection and provide research tools, along with access to live PCC nodes in research mode through its Apple Security Bounty Program. Apple said this will give security researchers the same level of visibility into the system as before.
More technical details are expected later this month at the Confidential Computing Summit. Apple also said it will update its PCC Security Guide and research program details later this year.
The expansion suggests Apple is trying to scale its privacy-focused AI strategy while leaning on outside infrastructure for heavier workloads. It also places Google and NVIDIA inside a system Apple says is still designed to keep privacy controls, approval authority, and transparency firmly under its own control.