Apple unveils AI cloud model built on Nvidia, Google tech

Apple Partners With Google and Nvidia to Power Its Most Advanced AI Cloud Model

Apple revealed on Monday at WWDC 2026 that it has built its most powerful AI model yet — AFM Cloud Pro — in collaboration with Google and Nvidia, running on Nvidia GPUs within Google Cloud infrastructure while extending Apple’s Private Cloud Compute privacy guarantees to third-party data centers for the first time.

A New Architecture for Apple Intelligence

During a technology session at the conference, Apple executives said the AFM Cloud Pro model is comparable to Google’s Gemini frontier models and will operate on Nvidia’s Blackwell B200 chips hosted in Google’s data centers. The announcement builds on a multi-year collaboration first disclosed in a joint statement from the two companies in January 2026, in which Apple and Google said “the next generation of Apple Foundation Models will be based on Google’s Gemini models and cloud technology”.theverge

Apple clarified that its Foundation Models are built on Apple’s own technology and code, with Google’s Gemini serving as a training and refinement resource through a process known as distillation rather than a direct integration. “Apple is NOT deploying Gemini models on user devices,” the company emphasized during the presentation. Instead, Apple uses a version of Google’s large Gemini model to train smaller models that can run locally on Apple devices, while more complex queries route to the cloud.wccftech

Privacy in Third-Party Data Centers

The expanded Private Cloud Compute infrastructure marks a departure from Apple’s prior approach, which relied exclusively on Apple Silicon servers. According to reporting from The Information, Apple approved the use of Nvidia’s confidential computing technology — a hardware-based feature that encrypts data and AI models during processing — to maintain its privacy standards on Google’s servers. Apple is expected to continue using the “Private Cloud Compute” branding for these features even though they will no longer run exclusively on Apple’s own hardware.9to5mac

Craig Federighi, Apple’s senior vice president of software engineering, also discussed the broader partnership with Google during the keynote, unveiling a more powerful on-device model alongside the cloud offering.seekingalpha

What Comes Next

The new Siri AI, powered by this architecture, is expected to launch in September with iOS 27. A public beta is set to begin next month, though Apple noted that some daily usage limitations will apply, with additional capacity available through iCloud+. Apple reportedly acquired 250 Nvidia NVL72 servers, each valued at approximately $4 million, to support the rollout.macdailynews