New details have emerged regarding the underlying technology for an upcoming, Gemini-powered version of Apple’s virtual assistant, Siri. Analysis indicates that Apple plans to leverage Nvidia’s advanced data center chips for certain processing tasks, potentially through a partnership with Google.
Confidential Compute for Enhanced Privacy
Sources familiar with the matter suggest that Apple will utilize Nvidia’s Blackwell B200 data center GPUs. A key feature of this integration will be Nvidia’s confidential compute technology. This hardware-based security system is designed to encrypt data while it is actively being processed by the GPUs, thereby safeguarding sensitive information even in cloud environments.
Nvidia states that this confidential compute feature is crucial for maintaining the privacy and integrity of AI models, enabling secure execution of demanding AI workloads at scale with minimal performance impact. This technology is available on Nvidia’s Rubin, Blackwell, and Hopper GPU platforms.
Strategic Shift and Integration Questions
This reported strategy marks a departure from Apple’s historical approach of maintaining tight control over the core components of its products. It remains unclear how Apple’s existing server system, known as Private Cloud Compute, will be integrated into the new Siri offering.
The agreement with Google is expected to route specific user queries to a licensed version of Google’s Gemini model, which will then run on these Nvidia chips hosted within Google Cloud. This move underscores Apple’s commitment to enhancing Siri’s capabilities while addressing privacy concerns through advanced technological solutions.
