The strategic reversal: vertically integrated Apple is now renting Gemini on Nvidia chips
For most of its modern history, Apple has narrated itself as the company that owns the stack — silicon, OS, services and now models. The Gemini-Siri deal punctures that narrative. Apple signed a multi-year contract for a custom ~1.2-trillion-parameter mixture-of-experts Gemini variant to power Siri's cloud features at a cost reported around $1 billion per year [1][3], and that model is being served alongside Nvidia silicon in Apple's new Siri infrastructure [11]. Apple's own statement to CNBC explicitly justifies the choice by saying Google's technology 'provides the most capable foundation for Apple Foundation Models' [4], which is corporate-speak for 'we did not get there in time'.
The X.com and YouTube communities have read this for exactly what it is. Commentators on developer-facing YouTube channels frame the deal as Apple effectively conceding the foundation-model race, while leak-and-rumor accounts on X highlight the optics of a vertically-integrated company renting compute and a model from a direct competitor. The strategic logic is still defensible — Ming-Chi Kuo argues the move will reshape the App Store by rewarding apps that expose rich App Intents and bypassing UI-locked apps [9]— but the philosophical break is real. Apple is now an integrator of best-of-breed AI rather than the sole author of it, and Tim Cook's last WWDC will be the keynote that makes that explicit.


