Apple is gearing up for its WWDC reveal of iOS 27 and a revamped Siri, but a new report today shows the AI engine behind the scenes relies heavily on Google’s Gemini model. The Information details how Apple will distill the massive Gemini network into a compact version that can run locally on iPhones and Macs, preserving the company’s on‑device narrative.

To bridge the gap where on‑device processing falls short, Apple will offload some queries to Google Cloud, running the full Gemini model on Nvidia’s confidential‑compute AI chips. Nvidia’s hardware encrypts data in transit, a trade‑off that slows latency but lets Apple promise privacy. Sources say the partnership with Nvidia was finalized only weeks ago, underscoring the urgency of a hybrid approach.

Apple is also scouting acquisitions to speed model compression, with Cambridge‑based Liquid AI mentioned as a likely target. While the private‑cloud branding will remain, the reliance on external GPU power signals a shift from purely internal compute. The move positions Apple to roll out AI features faster, but it also ties the ecosystem to Google’s cloud roadmap.