In a nutshell: Apple is implementing the new Siri generation in iOS 27 using Google’s Gemini models and leveraging Google Cloud for complex AI queries because its own Private Cloud Compute infrastructure lacks sufficient scalability.
Apple is offloading complex Siri queries for iOS 27 to Google Cloud, despite the company originally planning to use its own Private Cloud Compute infrastructure. Capacity bottlenecks in running Google’s Gemini models on Apple’s own servers force this deviation from the intended privacy strategy.
The upcoming iOS 27 generation, which Apple will present at WWDC 2026, integrates a fundamentally redesigned Siri with generative AI – based on Google’s Gemini model architecture. Apple’s original promise of primarily local data processing to protect privacy proves technically infeasible: full-featured contextual conversational AI with multi-turn dialogues and logical reasoning cannot be operated locally on smartphone hardware.
Apple is therefore implementing a hybrid system based on model distillation. The large Gemini model hosted in Google data centers serves as a “teacher” for a compact local model with an estimated 3 to 7 billion parameters. Through systematic training with question-answer pairs, the model learns to replicate the structures of the larger prototype. Subsequently, less relevant weights are removed (pruning) and mathematical precision is reduced (quantization). The resulting on-device model is optimized for Apple’s Neural Engine but proves insufficient for more complex queries.
Originally, Apple intended to process such complex queries via its own Private Cloud Compute infrastructure – dedicated AI servers with M-series chips that promise stateless data processing without persistent storage. However, internal reports reveal fundamental capacity bottlenecks: Apple was unable to stably and efficiently run Gemini’s uncompressed models on the M-chip clusters. Server capacity is insufficient for the massive data volume at iOS 27 launch. Consequently, complex voice queries are being routed directly to Google Cloud – a significant deviation from Apple’s original privacy strategy.
Source: www.it-daily.net · Published June 3, 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.2.9.