NVIDIA Confidential Computing is now powering a key part of Apple’s privacy-focused AI infrastructure, marking a notable collaboration between two technology giants. Announced at Apple’s annual Worldwide Developers Conference, the partnership brings NVIDIA GPUs into Apple’s Private Cloud Compute to handle secure, server-side inference for Apple Intelligence features.

A New Chapter for Private Cloud Compute

Apple’s Private Cloud Compute, known as PCC, is expanding beyond Apple’s own data centers and onto Google Cloud. As part of that expansion, NVIDIA GPUs equipped with Confidential Computing are now being used for confidential inference within the platform.

The move was unveiled during WWDC, Apple’s flagship event for developers around the world. Under the arrangement, NVIDIA GPUs will support server-side inference for Apple Foundation Models, which are custom-built by Apple and Google and draw on the technologies behind the Gemini family of models.

At the heart of the effort is a three-way collaboration. NVIDIA is working alongside Apple and Google to power some of the next generation of Apple Intelligence features, using NVIDIA Blackwell GPUs with Confidential Computing built directly into PCC’s hardware security architecture running on Google Cloud.

Why Confidential Computing Matters

As AI becomes more woven into everyday experiences, protecting user data has become a central challenge. NVIDIA Confidential Computing addresses this by adding a hardware-based security layer for accelerated AI workloads.

The technology safeguards data while it is actively being processed, a stage that has traditionally been difficult to secure. It does this by isolating workloads inside trusted execution environments and allowing systems to cryptographically confirm that the underlying infrastructure has not been tampered with before any sensitive data is ever sent to the server.

For everyday users, the implications are significant. With this approach, no one, not even the engineers who built the system, can access their data, chats, or conversations.

A Broader Shift in AI Infrastructure

Adopting NVIDIA Confidential Computing at this scale points to a larger trend reshaping how AI is delivered. Modern AI experiences increasingly blend on-device processing with cloud-based computation to complete their tasks.

That hybrid model creates a clear need: high-performance, server-side inference that still delivers strong privacy and security guarantees. Confidential Computing is designed to meet exactly that demand, letting powerful cloud systems handle sensitive work without compromising user trust.

How It Enforces Privacy and Trust

NVIDIA frames this technology as part of its broader commitment to trustworthy AI. Several core capabilities make that possible:

• Hardware-rooted trust, which helps confirm that workloads are running on genuine, untampered NVIDIA GPUs.
• Encrypted communication paths, protecting data as it travels between different components of the system.
• Remote attestation, allowing software to verify the platform’s security state before any sensitive data is released.
• Support for accelerated AI inference and training, so organizations can run privacy-sensitive workloads without sacrificing GPU performance.

Together, these features are becoming ever more important for AI services that must process sensitive information while preserving strong user privacy controls.

By embedding Confidential Computing into Apple’s Private Cloud Compute, NVIDIA, Apple, and Google are signaling that the future of AI need not force a trade-off between performance and privacy. Instead, the collaboration aims to deliver both at once, setting a notable example for how large-scale, privacy-conscious AI infrastructure can be built.