Apple has acknowledged that its most advanced cloud AI, the backbone for an upgraded Siri, will run on Nvidia's Blackwell B200 GPUs. The hardware is hosted within Google's cloud infrastructure, not purchased directly.
The key technology enabling the deal is Nvidia's confidential computing framework. It allows Apple to process sensitive AI workloads in Google's cloud without exposing user data to either Google or Nvidia. This privacy layer was essential for Apple's approval.
For Nvidia, securing Apple as a customer-even indirectly through Google Cloud-is a significant market signal, reinforcing its dominance in AI hardware. Google Cloud gains a major enterprise client, though it is simultaneously powering a direct competitor to its own Google Assistant.
Apple is not abandoning its own silicon. The company is developing an in-house data center inference chip, Project ACDC, expected in the second half of 2026. This project aims to reduce future dependence on third-party providers like Nvidia.
The strategy reveals a company bridging an immediate AI capability gap while pursuing a long-term proprietary solution.