Apple and Google join forces for a new Siri
Apple plans to use Nvidia Blackwell chips and Google cloud technologies to significantly expand the capabilities of its Siri voice assistant. Special attention is being given to protecting user data and ensuring hardware security during request processing.
Crius
A recent report states that Apple is considering using Nvidia Blackwell chips, which are already deployed in Google Cloud, to significantly upgrade its Siri voice assistant. This move is expected to greatly enhance Siri’s capabilities.
Some user requests to the new version of Siri on iPhone will be processed via Google Cloud, utilizing the licensed Google Gemini model and Nvidia Blackwell B200 data center GPUs. Apple has approved the use of Nvidia’s computing technologies, which provide data encryption during processing on these chips.
This collaboration is part of a broader agreement between Apple and Google aimed at creating a more powerful Siri with expanded features and accessibility options. The launch of the updated voice assistant is scheduled for September.
Apple plans to leverage a fleet of Google’s Nvidia Blackwell B200 data center chips and activate Nvidia’s confidential computing feature, which encrypts data as it is processed by the chips. Data encryption is a top priority for Nvidia, as it preserves the confidentiality and integrity of AI models running on Rubin, Blackwell, and Hopper GPUs, enabling secure execution of sensitive AI workloads in high-performance cloud and shared environments.
Nvidia’s flagship Blackwell GPU line is designed for large-scale artificial intelligence training and is being prepared for the next-generation Vera Rubin launch. The Blackwell B200 replaces the Hopper architecture, delivering significant improvements in performance, speed, memory bandwidth, and the ability to connect large numbers of GPUs.
Using this infrastructure will allow more demanding Gemini models to operate through Siri, handling complex tasks on iPhone, including multi-step reasoning, deep app integration, and improved contextual understanding. Privacy is a key focus: confidential computing creates a hardware-protected environment that encrypts both data and the model itself during processing.
As a result, even though search queries and commands pass through Google’s infrastructure and Nvidia hardware, the data remains protected from both companies, aligning with Apple’s long-standing privacy policy for iOS devices.
