Apple rebuilt Siri on Google’s AI and Nvidia’s chips, then spent WWDC explaining why that doesn’t break its…
The business move
Apple revealed at WWDC 2026 that it rebuilt Siri on a custom 1.2-trillion-parameter AI model based on Google’s Gemini architecture. The rebuilt Siri runs on Google Cloud infrastructure using Nvidia’s Blackwell B200 GPUs. This marks a notable shift in Apple’s AI strategy by outsourcing key AI inference workloads to the cloud of its largest tech rival.
Why it matters
Outsourcing AI to Google Cloud powered by Nvidia chips challenges Apple’s longtime privacy stance built on tightly controlled hardware and software. For years, Apple positioned privacy as a differentiator by limiting data exposure and running AI closer to the user device. Using Google’s cloud exposes Siri interaction data to external infrastructure, raising questions about data handling and compliance with Apple’s privacy promises. This move pressures Apple to justify how it maintains robust privacy despite relying on a major competitor’s platform.
Who gains and who gets squeezed
Google gains a substantial AI workload and revenue from Apple’s billions of devices, strengthening Google Cloud’s AI positioning and Nvidia’s GPU demand. Apple gains access to advanced generative AI models and high-performance chips without investing heavily in its own AI datacenters or chip fabrication. However, Apple risks erosion of consumer trust among privacy-conscious users and faces higher scrutiny from regulators and privacy experts scrutinizing outsourcing. Competitors emphasizing edge AI or proprietary chip stacks may capitalize on privacy-focused customer segments increasingly skeptical of cloud dependency.
What to watch next
Watch how Apple aligns this architecture with privacy guarantees in practice and communiqués, especially regarding data encryption and user metadata protections. Regulatory attention on cross-company AI data flows is likely to increase. Also watch Nvidia and Google as their chips and cloud services secure deeper critical workloads from tech rivals, reinforcing their market dominance. This architecture could prompt other hardware vendors to rethink how much AI runs internally versus in the cloud.
AI Quick Briefs Editorial Desk
