Apple May Use Google Cloud for Siri

This isn't the first time Apple has leaned on its rival's infrastructure; Google Cloud already provides Apple with online storage and computing power for training some of its in-house AI models. The potential expansion comes as Apple also maintains a significant cloud partnership with Amazon Web Services (AWS) for services like iCloud. The consideration to use Google's servers for Siri stems from the massive computational power required by large language models. Reports suggest Apple's own AI infrastructure is "beginning to decay" as it decommissions older Nvidia-powered servers, making it more reliant on third-party providers to handle the surge in demand expected from a smarter Siri. Google's key advantage lies in its custom-designed Tensor Processing Units (TPUs), hardware specifically optimized for the high-efficiency workloads of large AI models like Gemini. In contrast, Apple's Private Cloud Compute system runs on servers using its consumer device-focused silicon, which is not as well-suited for running large-scale AI models. Under the potential arrangement, Google would reportedly set up dedicated servers that adhere to Apple's stringent privacy requirements, a crucial step for integrating the infrastructure of two major competitors. This move follows a period where Apple's software chief, Craig Federighi, had previously vetoed using Google Cloud over security concerns, which were reportedly addressed by Google in 2023. This infrastructure decision is a critical part of Apple's broader AI strategy, which includes a major overhaul for Siri slated for 2026. The goal is to transform the assistant to better understand personal context from on-device data and handle more complex, multi-step tasks within apps.