Apple Taps NVIDIA Blackwell GPUs In Google Cloud To Power Next-Gen Siri AI
The alliance centers around the third generation Apple Foundation Models (AFM 3), which power next-gen Apple Intelligence features like the heavily upgraded Siri. While smaller AI tasks will continue to run locally on iPhone and Mac silicon, more complex reasoning and multi-step agent actions require some extra muscle. Thus, Apple developed AFM 3 Cloud Pro, a massive model optimized specifically for NVIDIA GPUs.
Historically, Apple has guarded its ecosystem by keeping data processing within its own custom-built data centers. However, the sheer demands of high-level generative AI have driven the company to look beyond its borders. By expanding its Private Cloud Compute (PCC) framework into third-party environments, Apple is utilizing Google Cloud as a hosting environment.
The primary obstacle in outsourcing cloud processing has always been data privacy, one area where Apple has built its brand identity. To resolve this, the partnership relies heavily on NVIDIA’s hardware-based confidential computing tech. Integrated directly into the Blackwell architecture, this system isolates sensitive user data in secure, trusted execution environments while it is being actively processed. Confidential compute acts as a digital vault, establishing a hardware-rooted trust and utilizing remote attestation to verify that the platform has not been tampered with before any personal data is transmitted.
For daily consumers, the background synergy ought to translate into a seamless, deeply integrated AI assistant. Behind the scenes, the framework guarantees that personal conversations, search queries, and files remain entirely secure. No one—including Google as the host, NVIDIA as the hardware manufacturer, or even Apple itself—can peek at the data.
As generative AI models grow increasingly sophisticated, the line between on-device processing and massive cloud computation is blurring, and if this alliance is any indicator, it could mean a change in how the AI infrastructure is built moving forward. Here, Apple is no longer pretending that every useful model can live entirely on an iPhone. It's betting that users will accept a more powerful assistant if the engineering behind it still feels locked down, inspectable, and hard to trust only in name.
Main image credit: NVIDIA
