Apple has spent years telling the world it builds everything itself. Its chips, its software, its servers. That whole philosophy is kind of the point of being Apple.
So when a report quietly revealed that the new Siri launching this September will run on Google’s fleet of Nvidia Blackwell chips, that detail deserves a second look.
Apple already has a cloud system built for exactly this kind of job. It’s called Private Cloud Compute, runs on Apple’s own Mac-series chips, and was designed from scratch to handle AI requests in a privacy-first way.
Apple even tried to get Google’s Gemini model running on it. The problem is it was too slow. So rather than ship something sluggish, Apple reportedly decided to just use Google’s infrastructure instead.
The Blackwell B200 chips Google is running are genuinely powerful hardware. Released as Nvidia’s follow-up to its Hopper generation, they were built from the ground up for large language model work, handling the kind of AI inference that a smarter Siri would need at serious scale.
Apple will lean on Nvidia’s confidential compute feature to keep user data encrypted while it’s being processed on Google’s servers, which is the main privacy concession Apple is making here.
Simpler queries will still be handled on your device. But anything that needs more intensive processing gets routed to the cloud, where it runs on Google’s Gemini models powered by those Blackwell chips.
Apple and Google formalized this arrangement earlier this year, and it covers the AI features Apple wants to highlight at WWDC 2026, which kicks off June 8.
Apple Intelligence launched back in 2024 with a reception that was, being generous, lukewarm. Several of the more ambitious Siri features were delayed, which eventually led to a $250 million class-action settlement.
WWDC this year is widely expected to be Apple’s attempt to reframe that whole story. Whether borrowing Google’s computing muscle helps Siri finally feel capable remains the actual question worth watching.
