Apple Partners with Google Gemini to Power Rebuilt Siri AI

Apple paying Google to power Siri is the most expensive "build vs. buy" decision in tech history -- and probably the right one. Apple gets best-in-class AI across 2B+ devices without burning $10B/year on training infrastructure.

TC
Trace Cohen
Early-stage VC & angel ยท Founder, New York Venture Partners
June 9, 2026
2 min read
KEY TAKEAWAYS FOR VCs & FOUNDERS
1

Google gets distribution no ad campaign could buy; Apple confirmed foundation model dev is a 5-player oligopoly

2

The deal validates that distribution + UX beats raw model capability -- a signal for every AI startup's positioning

TC
The VC Read ยท Trace's TakeTrace Cohen

Apple just told every AI-assistant startup the quiet part out loud: distribution wins, models rent. If the most capitalized company on earth would rather pay Google than build, what does that say about your plan to be the interface layer? The winners in applied AI won't own the model -- they'll own the workflow, the data, or the customer.

Apple is paying Google an estimated $3 billion or more annually to embed Gemini into the rebuilt Siri AI, announced at WWDC 2026. The partnership makes Google's models the intelligence layer behind Siri's natural language understanding, task execution, and multimodal capabilities across iPhone, iPad, Mac, Vision Pro, and Apple Watch. Apple's own internal AI efforts -- which reportedly consumed billions in R&D spend over three years -- were quietly shelved in favor of this deal.

The strategic calculus reveals a brutal truth about the AI market: even a $3 trillion company with the world's best custom silicon team concluded it couldn't compete at the foundation model layer. Apple's M-series chips are extraordinary for inference, but training frontier models requires a different kind of infrastructure -- massive GPU clusters, petabytes of curated training data, and teams of researchers that take years to assemble. Apple tried and failed. The fact that they chose Google over OpenAI or Anthropic likely came down to integration depth: Gemini's multimodal capabilities (text, image, video, code) across Google's existing service ecosystem made it the most versatile option for Siri's broad use cases.

โ€œGoogle gets distribution no ad campaign could buy; Apple confirmed foundation model dev is a 5-player oligopolyโ€

For Google, this is the distribution deal of a generation. Gemini will run on 2 billion+ active Apple devices without Google having to build, market, or support a single piece of hardware. It's the inverse of the Google Search default deal that's been under DOJ antitrust scrutiny -- except Apple is paying Google this time, making the antitrust optics cleaner. Google's AI division can now credibly claim the largest deployment footprint of any foundation model provider, surpassing even OpenAI's ChatGPT user base.

The startup implications are severe. Every company building a general-purpose AI assistant -- whether hardware (Humane, Rabbit) or software -- just watched their total addressable market shrink dramatically. The remaining opportunity is vertical: domain-specific AI that knows more about your specific workflow than Siri + Gemini ever will. If your AI startup's pitch includes the phrase "AI assistant," it's time to pivot to "AI expert in [specific domain]."

Originally reported by Tom's Guide. Analysis and editorial commentary by Value Add Pulse.

โ† Back to Pulse