Claim of Google in Apple's Next-Generation Siri Infrastructure

Infrastructure Talks Between Apple and Google

According to a new Bloomberg-sourced report, Apple is holding infrastructure talks with Google for future versions of its digital assistant Siri. The claims indicate that Apple may use Google’s servers equipped with Tensor Processing Units (TPUs), designed specifically for AI operations, instead of its own developed “Private Cloud Compute” architecture. It is stated that this potential collaboration is being considered especially for a much more capable chatbot planned to be introduced with iOS 27.

It is stated that the new large language model (LLM), developed within Apple under the code name “Apple Foundation Models v11,” requires a level of processing power that will push the limits of current Apple Silicon-based servers. To offer this AI system globally, which is intended to compete with Google’s Gemini models, Google’s already deployed and scalable TPU infrastructure may have become a technical necessity.

Technical Details and Privacy Protocols

• Infrastructure Shift: Apple’s move toward Google TPUs by stepping outside its own hardware ecosystem reveals the rising compute costs and hardware requirements of AI models.
• Timing: Since the upcoming iOS 26 interim updates (26.4, etc.) will continue to run on current Apple servers, the main major architectural change is expected to take place in the iOS 27 period.
• Data Security Model: Apple may adapt to Siri the strategy of hosting data on third-party servers (Google Cloud, AWS) by encrypting it, as in iCloud services. In this scenario, control of the encryption keys would remain entirely with Apple, preventing Google from reading the data.

Although this development means Apple is departing from its “everything in-house” (on-device or Private Cloud) approach, a transition to a hybrid infrastructure model appears inevitable due to the nature of cloud-based generative AI services.