Apple expands Private Cloud Compute to Google Cloud and NVIDIA GPUs
Apple is extending its Private Cloud Compute infrastructure by utilizing NVIDIA Blackwell GPUs and Google Cloud's confidential computing architecture. The deployment integrates Intel TDX and Google Titan chips to maintain cryptographic privacy for high-demand AI workloads.
Key Takeaways
- Apple deployed its AFM 3 Cloud Pro model on Google Cloud using NVIDIA Blackwell GPUs for high-demand AI tasks.
- Hardware security layer combines NVIDIA Confidential Computing, Intel TDX memory protection, and Google Titan hardware root of trust.
- The 2026 expansion allows Apple to offload 'agentic tool-use' and complex reasoning workloads while maintaining existing privacy models.
- Apple retains full software control by cryptographically approving only verified binaries for deployment on third-party infrastructure.
- Deployment is currently in a summer preview period and will gradually implement the full security architecture through late 2026.
Why It Matters
This move signals that even the industry's most vertically integrated giants cannot scale the next generation of generative AI using in-house silicon alone. By anchoring privacy in a multi-vendor hardware stack rather than just Apple-owned data centers, the company creates a scalable blueprint for 'confidential inference' that balances massive compute requirements with rigid data sovereignty. For the streaming and broader digital ecosystem, it validates NVIDIA Blackwell as the foundational accelerator for high-security cloud workloads. Watch for Apple to release its public PCC binaries later this year, which will permit independent researchers to verify whether third-party cloud hosting truly maintains Apple's absolute privacy claims.
Additional Context
The deployment of Private Cloud Compute (PCC) on Google Cloud marks a strategic shift from Apple’s original 2024 launch, which relied exclusively on Apple silicon within proprietary facilities. According to Apple Security Research (June 2026), the current extension to Google Cloud was prompted by the extreme compute demands of the Apple Foundation Model (AFM) 3 Cloud Pro. While smaller models in the AFM 3 family run on iPhones or Apple’s own servers, the 'Pro' variant required NVIDIA’s advanced Blackwell architecture to handle complex reasoning at the speeds consumers expect for a reworked Siri. This partnership also formalizes a deeper technical alignment between Apple and Google. Per reports from 9to5Mac and MacRumors (June 2026), the AFM 3 models were custom-built in collaboration with Google, using training technologies from the Gemini model family. Although Apple maintains that its models are distinct from Gemini, the underlying infrastructure on Google Cloud suggests a pragmatic move to use Google's existing AI Hypercomputer capacity rather than attempting to build out equivalent Blackwell-scale clusters in-house. Beyond the consumer applications, the move reinforces the commercial viability of confidential computing in a crowded hyperscale market. Per Gartner (June 2026), confidential computing is a core architecture trend for the year, projected to secure a significant portion of AI processing in untrusted environments by 2029. By being the first to integrate NVIDIA Blackwell, Intel TDX, and Google Titan into a single production pipeline, Apple and Google are setting a high baseline for secure AI infrastructure that enterprise buyers are likely to monitor for benchmarking their own data-sensitive deployments.
Read full article at astutegroup.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source