Google Announces Private AI Compute Platform

Google announced Private AI Compute on November 12, 2025, a new cloud-based platform designed to harness the advanced capabilities of its Gemini AI models while maintaining the privacy levels typically associated with on-device processing. The initiative targets users of Pixel smartphones and other compatible devices, addressing the computational limitations of local hardware by securely offloading complex AI tasks to the cloud.

The platform enables AI tools to process sensitive user data in a fortified environment, isolated from general cloud operations. According to Google's official blog post, data is handled within specialized hardware that prevents access by external parties or even Google employees. This is achieved through a multi-layered architecture featuring the company's custom Tensor Processing Units, or TPUs, equipped with Titanium Intelligence Enclaves for hardware-based isolation. An AMD-based Trusted Execution Environment encrypts and segregates memory, while remote attestation and end-to-end encryption secure connections between user devices and the cloud. "Private AI Compute is like running AI on your device, but with the power of our best cloud models," the...