How Apple rebuilt Siri with Google's AI
Apple (AAPL) unveiled an AI-powered Siri at WWDC, saying it will launch as a beta later this year. Apple said Siri AI can use app context to surface answers without opening other apps. Apple uses Google’s Gemini models under a January agreement, according to Craig Federighi, while training its own AFM on proprietary data and refining with Gemini outputs. Apple also expanded Private Cloud Compute with Nvidia chips in Google Cloud.
Siri AI readiness and the disclosed compute/model sourcing (AFM + Google Gemini outputs; Nvidia-powered PCC in Google Cloud) are a tangible product-technology milestone for Apple’s AI platform.
Apple unveiled AI-powered Siri at WWDC and said it will launch as a beta later this year, powered by Apple Intelligence.
Near-term sentiment tailwind into the fall beta launch; upside skew if demos translate into user adoption and developer integration.
Background
Apple had delayed its reimagined Siri for about two years and previously previewed AI-enhanced Siri and Apple Intelligence at WWDC 2024.
Why it matters
The rebuilt Siri AI is positioned as a context-aware assistant that can surface app and web content, with Apple Frontier Models trained/refined using Gemini outputs and cloud compute extended to Nvidia chips in Google Cloud.
Market relevance
Traders may reprice Apple’s AI execution risk and the AI compute/model partnership ecosystem ahead of the fall beta.
Market effects
Reinforces the trend of major consumer platforms integrating third-party frontier models and using cloud GPU compute to scale on-device AI assistants.
Primarily US mega-cap sentiment (Apple/Google/Nvidia) with spillover to US AI infrastructure and consumer tech narratives.
Highlights cross-company AI supply chains (model providers + cloud compute + device ecosystem) that can influence global AI capex expectations.
Alternative perspectives
A beta launch later this year may still face quality, latency, and privacy/UX hurdles; early demos may not translate into sustained engagement or monetization.
The article emphasizes model sourcing and compute plumbing but omits performance benchmarks, user privacy tradeoffs in practice, and whether Apple can match competitors’ assistant reliability at scale.
Key entities
- companyApple
Launched AI-powered Siri at WWDC and plans a beta later this year, using Apple Intelligence and AFM.
- companyGoogle
Gemini models provide outputs used to refine Apple’s AFM; Apple also uses Google Search for Gemini grounding but says it does not use Google Search directly.
- companyNvidia
Nvidia chips are used to extend Apple’s Private Cloud Compute in Google Cloud for the high-end ARM Cloud Pro model.
