New information continues to emerge regarding Apple’s artificial intelligence plans taking shape before WWDC. A new news published by The Information shares remarkable details about the infrastructure that the company plans to use in the Gemini-powered Siri experience. According to the report, Apple will use the Gemini model running over Google Cloud to process some Siri queries. In addition, it is stated that Nvidia’s new generation data center chips will be used in some of these transactions.
In the news previously published by The Information, it was stated that within the scope of the agreement made between Apple and Google, some user queries directed to the new Siri version will be processed by a licensed Gemini model running on Google Cloud. In the same news, it was stated that Apple approved the use of a privacy technology developed by Nvidia in this process. This indicated that Apple would benefit from Nvidia hardware for at least a certain portion of its AI workloads on Google Cloud.
The new report clarifies more clearly which hardware this infrastructure will work with. Accordingly, Apple will use Nvidia Blackwell B200 chips located in Google’s data centers. It is stated that the company will also enable the security feature that Nvidia calls “confidential compute”. Thus, user data can be protected encrypted not only during transmission but also during artificial intelligence operations.
Nvidia Blackwell B200 and secure transaction technology will be used in Siri infrastructure
Nvidia Blackwell B200 is among the most powerful data center GPUs the company has developed for large-scale AI training and inference processing. These chips, built on Blackwell architecture, are the continuation of Nvidia’s previous generation Hopper platform. The company states that the Blackwell family offers significant performance increases, especially in running large language models.
Bringing significant improvements in areas such as memory bandwidth, multi-GPU scaling and AI inference performance, the Blackwell platform is designed to support models at the trillion parameter level. The fact that Google uses this hardware in its own cloud infrastructure makes it possible to run advanced artificial intelligence models such as Gemini with high performance. Apple’s use of this infrastructure may contribute to the company’s ability to deploy productive artificial intelligence features more quickly.
On the other hand, Nvidia’s confidential compute technology is defined as a hardware-based security system that aims to protect data during processing. While traditional security methods mostly protect data during the storage or transfer stages, this technology ensures that information in the active transaction process remains secure against unauthorized access. According to Nvidia, the system can protect the confidentiality and integrity of AI models running on Blackwell, Hopper and future Rubin GPU platforms. In addition, it provides high security for sensitive artificial intelligence workloads running in shared or cloud-based environments, while keeping performance loss to a minimum.
It is known that Apple has preferred to develop its hardware, software and service layers under its own control as much as possible for many years. Despite this, it is noteworthy that the company turned to Google Cloud and Nvidia technologies for some of Siri’s artificial intelligence operations. The Information also emphasizes that this approach differs to a certain extent from Apple’s understanding of directly controlling critical technologies.
However, the role of Apple’s previously announced Private Cloud Compute infrastructure in the new Siri experience is not yet clear. This system, introduced last year, aimed to perform cloud-side operations of Apple Intelligence features on the company’s own servers. The new report does not provide definitive information on how Private Cloud Compute and Google Cloud-based Gemini integration will work together. The announcements to be made at the WWDC event are expected to reveal more clearly how Apple will establish the balance between privacy policies and third-party artificial intelligence services.