Overview
- The Information, as relayed by IT Home, reports Apple is allowing Google to run servers inside Apple facilities to handle a new Siri powered by Gemini.
- Roughly 90% of Apple’s Private Cloud Compute capacity is said to be idle, with only about 10% deployed in data centers and some AI servers still sitting in warehouses.
- Sources cite fragmented internal AI stacks and servers built with consumer‑oriented chips not optimized for large models as key constraints, following years of hesitance to expand cloud infrastructure.
- Apple is reportedly planning a major Siri update this year using a custom 1.2‑trillion‑parameter Gemini model (Foundation Models v10) and a separate Siri chatbot in iOS 27 that would run on Apple‑owned systems built on Google TPUs and cloud.
- For longer‑term self‑reliance, Apple is said to be developing an in‑house AI server ASIC codenamed Baltra with Broadcom and TSMC using a 3nm N3E chiplet design.