Apple’s next Siri overhaul may rely on Google Cloud, Gemini and Nvidia hardware working together behind the scenes, according to a new report from The Information that adds more detail to the company’s privacy pitch and cloud strategy.
Apple’s Gemini-powered Siri may run on Google’s Nvidia chips
The report builds on earlier claims that some Siri requests will be handled in Google Cloud using a licensed version of Google’s Gemini model. The latest details say Apple will tap into Google’s fleet of Nvidia Blackwell B200 data center chips for at least some of that processing.
Apple has also reportedly approved the use of Nvidia’s confidential compute technology in that setup. That feature encrypts data while it is being processed on the chips, adding another layer of protection for cloud-based AI tasks.
In other words, some Siri queries could be routed away from Apple’s own infrastructure and handled on Google-hosted systems that combine Gemini software with Nvidia hardware and security tools.
What Nvidia’s Blackwell B200 brings to the table
The Blackwell B200 is one of Nvidia’s flagship data center GPUs, built for large-scale AI training and inference. It follows Nvidia’s Hopper architecture and is designed to handle very large models, including trillion-parameter systems, with improved inference performance, memory bandwidth and multi-GPU scaling.
That makes the chip a logical fit for cloud AI workloads, especially when companies want to run demanding models without relying entirely on their own private server hardware.
How confidential compute is supposed to work
Nvidia describes confidential compute as a hardware-based security feature that protects data while it is actively being processed by GPUs. According to Nvidia, it “preserves the confidentiality and integrity of AI models deployed on Rubin, Blackwell, and Hopper GPUs,” and allows “sensitive AI workloads to run securely at scale with near-native performance, even in shared or cloud environments.”
The new report says Apple will enable that capability in Google Cloud as part of the Siri arrangement. That detail is notable because Apple has repeatedly emphasized privacy as a core part of its AI story.
Privacy messaging remains central
Apple has been stressing privacy assurances amid its growing reliance on outside AI partners. The Information previously reported that the company was expected to use Nvidia chips through Google Cloud for some Siri queries, and the newest report suggests Apple has now formally approved Nvidia’s privacy technology for that environment.
The move could help Apple reassure users that cloud-based processing will still be secured, even if some requests leave Apple-controlled systems. At the same time, it underscores how much Apple’s AI plans may depend on external infrastructure.
How this fits with Apple’s usual strategy
The Information says Apple’s approach “diverges from [the company’s] strategy of attempting to control all the critical ingredients to its products.” That is a significant point for a company that has long preferred to manage its own hardware, software and services stack as tightly as possible.
The report also says it is “unclear how Apple’s previously launched server system, called Private Cloud Compute, will fit into the upcoming Siri product launch.” That leaves open an important question: whether Apple will use its own cloud systems for some requests and Google’s infrastructure for others, or whether the new Siri experience will rely more heavily on partner-hosted AI than expected.
What remains unanswered
Even with these new details, there are still major gaps in the picture. The report does not spell out which Siri requests will run in Google Cloud, how Apple will decide when to route queries there, or how much of the assistant’s traffic will depend on Nvidia-powered servers.
It also does not explain how Apple’s Private Cloud Compute system will interact with the new setup, or whether the Gemini-based Siri experience will be visible to users in any meaningful way. For now, the reports point to a hybrid architecture that blends Apple’s privacy branding with Google’s AI models and Nvidia’s latest data center chips.
Why the hardware choice matters
Using Nvidia’s Blackwell B200 chips suggests Apple is aiming for serious AI capacity rather than a small-scale experiment. The choice also highlights how cloud AI deployments increasingly depend on specialized hardware, not just model access.
For Apple, the hardware and security layer matter as much as the model itself. If Siri is going to lean on a licensed Gemini version in Google Cloud, then the company appears to be trying to preserve trust by adding Nvidia’s encrypted processing tools on top.
That could prove important as Apple prepares to explain its AI direction to users, developers and investors. A Siri that uses outside models is one thing; a Siri that uses outside models plus outside cloud infrastructure is another. The confidential compute layer may be Apple’s answer to the obvious privacy concern.
Still, The Information’s reporting suggests Apple is willing to make a pragmatic tradeoff: using partner technology where it thinks it can improve Siri, while attempting to retain as much user confidence as possible through encryption and security controls.
Source: Original report
Was this helpful?
Last Modified: July 7, 2026 at 9:36 pm
3 views
