Jun 10, 2026 11:34:00

Apple's privacy-focused AI servers run on NVIDIA GPUs.

Apple is building a cloud system called '

Private Cloud Compute (PCC) ' that runs AI on high-performance servers while protecting user privacy. It has now been revealed that NVIDIA's confidentiality-focused system, ' Confidential Computing ,' will be adopted for this PCC.

NVIDIA Confidential Computing to Help Expand Apple's Private Cloud Compute | NVIDIA Blog
https://blogs.nvidia.com/blog/nvidia-confidential-computing-apple-private-cloud-compute/

Expanding Private Cloud Compute - Apple Security Research

https://security.apple.com/blog/expanding-pcc/

PCC is a system designed by Apple specifically for private AI processing, which processes user data in the cloud while protecting users' personal data with end-to-end encryption.

Confidential Computing is described as a tool that 'safely deploys cutting-edge proprietary models in on-premises or national environments while protecting the intellectual property rights of the models, and accelerates computing performance.'

On June 8, 2026, Apple announced its next-generation Apple Intelligence , based on Google's AI 'Gemini,' and revealed that it would expand its PCC (Personal Computer Center) beyond Apple's own data centers.

According to Apple, Apple collaborated with Google to build ' Apple Foundation Models, ' the underlying models for Apple Intelligence features, leveraging the technology behind Gemini. These models cover multiple operations, from on-device processing to cloud-based processing, but for the most demanding tasks, such as agent-based tool usage and complex inference, they extend PCC to the Google Cloud system using NVIDIA GPUs.

Apple stated, 'Initially built on Apple Silicon and leveraging world-class security technology, PCC set a new standard for AI privacy in the cloud. Since then, the entire industry has been working to provide confidential inference systems that can achieve the same level of security as PCC, but until today, these systems have never been integrated into a globally operational, end-to-end confidential inference system. PCC on Google Cloud has just done that.'

Note that PCC on Google Cloud runs on NVIDIA Blackwell GPUs.

NVIDIA stated, 'The adoption of Confidential Computing on this scale signifies a major shift across the entire AI infrastructure. As AI experiences combine on-device and cloud-based processing, there is a growing need to deliver high-performance server-side inference while maintaining strong privacy and security assurances.'

- Continued
What's so amazing about the new AI 'Apple Foundation Models'? A groundbreaking mechanism that runs the 20 billion parameter multimodal model 'AFM 3 Core Advanced' on an iPhone - GIGAZINE