Apple's Gemini Siri Runs on Nvidia Blackwell Chips
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Apple will send some queries from a new version of Siri to Google Cloud, where they will be processed by a licensed version of Google's Gemini model.
- The new Siri will run on Nvidia Blackwell B200 data center chips from Google's fleet — Nvidia's flagship AI GPU and the successor to the Hopper architecture.
- Apple approved Nvidia's confidential compute feature for the deployment, which encrypts user data while it is actively being processed on the GPUs.
- Nvidia's confidential compute works across its Rubin, Blackwell, and Hopper GPU lines, letting sensitive AI workloads run securely in shared or cloud environments at near-native performance.
- The Information notes the arrangement diverges from Apple's longstanding strategy of controlling all critical ingredients to its products.
- Apple's Private Cloud Compute server system — the company’s own privacy-focused AI infrastructure — has no clear role in the upcoming Siri launch.
- The revamped Siri is expected to debut in September with Gemini-powered capabilities.
Why it matters: Apple built Private Cloud Compute specifically to keep Siri queries in its own stack — a core piece of its privacy marketing. Routing Siri through Google Cloud and Nvidia silicon, even with hardware-level encryption, marks a pragmatic concession that competing in frontier AI requires infrastructure Apple doesn't have. September's launch will be the first major test of whether customers accept Siri as a Google-and-Nvidia-backed product.
Ask SkimNews
