Apple's Overhauled Siri to Run on Nvidia's Blackwell Chips, Powered by Google Gemini

Summary

Apple's upcoming overhauled version of Siri, set to launch in September, will rely on Nvidia's Blackwell B200 data center chips for cloud-based processing, according to The Information. When Siri queries require cloud-based computation, the system will fall back on Google's Gemini large language models, with user data encrypted using Nvidia's hardware-based confidential compute feature. This arrangement marks a significant departure from Apple's historical strategy of controlling all critical components of its products.

The decision reflects practical constraints in Apple's infrastructure. The company reportedly attempted to run a modified version of Gemini on its own Private Cloud Compute system—which runs on Apple's custom Mac-series chips—but found the model's performance was too slow for production use. Instead, Apple will outsource this critical cloud processing to Google's infrastructure powered by Blackwell chips, which are optimized specifically for large language models and offer dramatic speed improvements over previous generations.

This partnership arrangement underscores a broader industry trend: even vertical-integrated technology giants must sometimes rely on specialized partners to deliver AI performance at scale. While Apple will maintain the Private Cloud Compute branding for on-device processing, the cloud fallback will depend entirely on external infrastructure operated by competitors.

• This represents a strategic compromise: Apple is outsourcing critical AI infrastructure to Google and Nvidia despite its preference for vertical integration