Apple Pays Nvidia to Run Siri in Google Cloud — The Privacy Paradox You Were Not Told About

Fri, 05 Jun 2026 00:00:00 +0000

In late May 2026, The Information revealed what many in the AI industry had suspected: Apple has agreed to use Nvidia chips inside Google Cloud to run parts of the new, Gemini-powered Siri. The arrangement is part of a broader Apple-Google AI deal first reported in March 2026.
The setup is structurally bizarre. Apple — the company that has marketed privacy as its competitive moat for a decade — is now shipping user Siri queries to a third-party cloud (Google) running on a third-party AI accelerator (Nvidia). The “private cloud” promise of Apple Intelligence 1.0 (2024) was that queries never leave Apple-controlled infrastructure. The 2.0 Siri breaks that promise.
This is not a temporary engineering compromise. According to three former Apple engineers speaking to The Information, the Apple-Google-Nvidia architecture is the only path Apple has to ship a competitive agentic Siri in iOS 27 (autumn 2026). Below — the architecture, the legal exposure under EU AI Act and GDPR, the financial mechanics, and what it means for the 1.5 billion Apple devices that will be running this by 2027.

What was announced #

The May 2026 reports describe a three-party arrangement: