Particle.news
Download on the App Store

Apple Unveils Siri AI That Combines On‑Device Models With a Private Cloud Layer

The company says the assistant keeps user data private while routing heavy tasks to a confidential cloud run on Google infrastructure with Nvidia GPUs.

Overview

  • Apple announced Siri AI at WWDC on Monday and released developer betas with a public beta planned later this year and a wider consumer rollout in autumn.
  • The system pairs a second‑generation Apple Foundation Model that runs on devices with a Private Cloud Compute layer that offloads heavier inference to Google Cloud using Nvidia Blackwell GPUs under confidential compute protections.
  • Apple frames the product as privacy‑first and says user data is used only to execute requests, even as the architecture relies on external cloud and GPU partners to meet real‑time needs.
  • Early hands‑on tests and developer reports show clear capability gains over legacy Siri but note practical limits including slow on‑device indexing that can take days, limited third‑party app control, no persistent memory between sessions, and occasional accuracy gaps.
  • Analysts say Siri AI clarifies Apple’s path to monetizing AI through iCloud+ limits and higher‑end device features, but they remain unsure it will meaningfully drive upgrades or services revenue.