Apple Expands Private Cloud Compute to Google Cloud Infrastructure

Apple has announced the expansion of its Private Cloud Compute (PCC) infrastructure beyond its own data centers to include Google Cloud systems [1]. This strategic shift allows Apple to utilize NVIDIA GPUs to support its most advanced Apple Intelligence workloads while maintaining its strict privacy and security standards [1].

Key takeaways

• Apple is partnering with Google and NVIDIA to run complex Apple Intelligence tasks on Google Cloud servers [1].
• The new "AFM Cloud Pro" model is designed for agentic tool use and complex reasoning, utilizing NVIDIA GPUs in Google’s cloud [2].
• Apple maintains that it uses none of Google’s Gemini models, client-side code, or search infrastructure for its own systems [2].
• macOS 27, officially named macOS Golden Gate, introduces "Siri AI" and improved system performance [4].
• Apple’s Private Cloud Compute on Google Cloud is currently in beta and will be subject to public inspection and security bounty programs [1].

Integrating Third-Party Cloud for AI Processing

To support the most capable model in its new Apple Foundation Model (AFM) family, AFM Cloud Pro, Apple required additional computing power beyond its own silicon servers [2]. By extending PCC to Google Cloud, Apple can leverage NVIDIA GPUs configured with "ambiguous confidential compute" technology, which ensures that the cloud provider cannot access the data being processed [2]. Apple emphasizes that it retains full control over the PCC software, and its devices will only trust binaries that have been cryptographically approved by Apple [1].

While Apple is utilizing Google’s hardware infrastructure, the company clarified that its AI systems remain distinct from Google’s own technology. Apple executives stated that they do not use Google’s Gemini models, client-side code, or Google Search infrastructure [2]. Instead, Apple’s models were custom-built for Apple Silicon and refined using outputs from Gemini frontier models through a process of distillation [2]. To ensure transparency, Apple plans to provide public research tooling and access to live PCC nodes, allowing third-party researchers to verify that user data is not stored or accessed [1].

macOS Golden Gate and the Future of Siri

Alongside the cloud infrastructure updates, Apple unveiled macOS 27, titled macOS Golden Gate [4]. This release focuses on performance improvements and design refinements, such as a unified toolbar and edge-to-edge sidebars [4]. A significant addition to the operating system is "Siri AI," a conversational assistant integrated into Spotlight that provides personal context and on-screen awareness [4]. This update marks a transition for the platform, as macOS 27 will be compatible exclusively with Apple silicon Macs [3].

Why it matters

The expansion of PCC to Google Cloud represents a significant evolution in how Apple balances the need for massive computational power with its commitment to user privacy. By implementing a cryptographically verifiable ledger and hardware-level security like Intel’s TDX and Google’s Titan chip, Apple aims to extend its "no-privileged-access" guarantees to third-party environments [1]. As Apple prepares for a leadership transition—with John Ternus set to take over as CEO from Tim Cook on September 1—these AI-centric infrastructure and software updates define the company's technical roadmap for the coming year [3]. Users can expect the new Siri features and macOS Golden Gate to arrive later this year, though some features may initially be gated behind a waitlist as the company continues to refine the technology [3].