What Apple’s Siri Pivot Really Means
Apple’s shift from its own AI server infrastructure to Google Cloud and Nvidia Blackwell chips for the Siri redesign is a strategic reversal where performance, scale, and time-to-market have been prioritized above Apple’s usual end-to-end control of hardware, software, and cloud services, raising new questions about Apple Intelligence privacy and long‑term AI independence. The new Siri, expected around September 2026 with iOS 27, is described as a full rebuild that turns the assistant into a more conversational, context-aware AI system. According to The Tech Portal, Apple will use Google’s Gemini models alongside on-device Apple Intelligence for a hybrid design. Simple tasks will stay local on Apple Silicon, while complex requests move to the cloud. This two-layer approach is Apple’s answer to rivals that already offer powerful AI chat tools, but it comes at the cost of relying on third-party AI server infrastructure.
From Apple-Only Servers to Google Cloud and Blackwell
For years, Apple insisted Apple Intelligence would run on Apple Silicon and, when needed, on Apple’s own servers under its Private Cloud Compute system. AppleInsider notes that Craig Federighi said in 2024 it was “essential for privacy and security” that cloud processing use only Apple servers. That plan has changed. Apple reportedly tried running a version of Google Gemini inside Private Cloud Compute, but the setup was too slow to be usable. Instead, the company is turning to Google Cloud infrastructure powered by Nvidia’s new Blackwell B200 GPUs to handle demanding Siri features like document summarization and multi‑app automation. These Nvidia Blackwell chips are designed for large-scale AI workloads, letting Siri tap far larger models than Apple’s in-house cloud systems can support today. The result is a marked move away from Apple’s traditional vertical integration in AI server infrastructure.
Gemini in the Cloud, Apple Intelligence on the Device
The redesigned Siri follows a hybrid model: on-device Apple Intelligence for quick, private tasks, and cloud-based Gemini for advanced language work. The Tech Portal reports that Apple is paying around USD 1 billion (approx. RM4.6 billion) per year for access to a customized Gemini model with about 1.2 trillion parameters, far larger than Apple’s own cloud models at roughly 150 billion parameters. This size gap explains why Apple is turning outward. On iPhone and other devices, users will see Siri gain deeper awareness of personal context across Mail, Messages, Calendar, Photos, and Notes, and they will be able to run multi-step commands through apps. When a request exceeds what local Apple Silicon can handle, it will be sent to Google Cloud infrastructure running Gemini and Nvidia Blackwell chips, with Apple also using model distillation so its smaller models can learn from Gemini’s responses over time.
Can Apple Keep Its Privacy Promise on Third-Party Clouds?
Apple has built its brand around strong privacy, including the idea that Apple Intelligence and Siri requests would either stay on device or go only to Apple’s own secure servers. Using Google Cloud infrastructure complicates that message. AppleInsider reports that Apple will enable Nvidia’s confidential computing on Blackwell B200 GPUs, so data stays encrypted even while processed. Apple’s existing rules also say prompts sent to the cloud are not retained for training. Yet Apple’s earlier public stance—that sensitive AI tasks needed Apple‑controlled servers—makes this pivot notable. The company will need to explain how Apple Intelligence privacy is preserved when requests pass through Google’s data centers, even with encryption. The move shows that, when forced to choose, Apple has favored speed and capability over pure isolation, betting that technical safeguards are enough to reassure users.
What the Siri Deal Signals for Apple’s AI Ambitions
The decision to abandon a fully in-house AI server infrastructure for Siri sends a clear signal: Apple is willing to trade some control for competitive AI performance. AppleInsider points out that it is unusual for Apple not to control “the whole stack, from its own servers to its own software.” Instead, the refreshed Siri will lean on Google Gemini, Google Cloud, and Nvidia Blackwell chips, even as Apple continues to develop its own models and on-device Apple Intelligence features. In the near term, this strategy could let Apple catch up to rivals in conversational AI and complex automation. In the long term, it raises open questions: can Apple reduce dependence on Google and Nvidia as its own models mature, and will users accept privacy guarantees that rely on confidential computing rather than Apple‑owned servers? The answers will shape Apple’s AI narrative beyond this Siri redesign.
