Apple has advanced plans to use Nvidia’s Blackwell B200 chips in Google data centers for its long-delayed Siri overhaul.
Summary
- Apple has reportedly advanced plans to use Nvidia Blackwell B200 chips for its delayed Siri AI upgrade.
- The chips will be hosted in Google data centers, according to The Information’s latest report.
- Apple tested a Gemini-based model on Private Cloud Compute, but performance was reportedly too slow.
- The company may use Nvidia confidential computing to protect cloud-based Siri requests during processing.
The Information reported that Apple plans to route cloud-based Siri requests through Google’s Nvidia-powered infrastructure after tests on its own Private Cloud Compute system failed to deliver the speed needed for the upgraded assistant. The report said Apple tried to run a modified version of Google’s Gemini model on its PCC servers, but the performance was too slow for practical use.
The expected announcement could come during Apple’s Worldwide Developers Conference keynote on June 8, where the company is also expected to preview iOS 27 and major Apple Intelligence updates. Yahoo Tech reported that the refreshed Siri could arrive for users in September alongside the iOS 27 release.
Apple turns to Google-hosted Nvidia chips
According to The Information, Apple will use Nvidia Blackwell B200 processors hosted inside Google’s data centers to support cloud-based Siri tasks. The move would place part of Siri’s AI workload on outside infrastructure, even though Apple has long preferred to control its hardware, software, and cloud systems.
At WWDC 2024, Apple senior vice president Craig Federighi said cloud-based Apple Intelligence tasks would run on Apple’s own servers through Private Cloud Compute. Apple described PCC as a privacy-focused system built on Mac-series chips that could process requests without storing user data or exposing it to outside parties.
Two years later, The Information’s report suggests Apple needs more powerful AI infrastructure to deliver the Siri features it first promised in 2024.
Privacy remains central to the rollout
MacRumors reported that Apple may keep the Private Cloud Compute name even if some Siri requests move through Google-hosted Nvidia hardware. The report said Apple plans to use Nvidia’s confidential computing technology to protect user data while it is processed in the cloud.
Nvidia says its confidential computing tools encrypt data during processing on supported hardware. For Apple, the feature could help answer privacy questions created by the use of third-party data centers.
The final structure of the system remains unclear, according to MacRumors, because Apple has not publicly explained how PCC branding would apply to Nvidia chips hosted by Google.
Siri overhaul follows years of delays
Yahoo Tech reported that the Siri update will be the assistant’s largest redesign since its 2011 launch. Apple first previewed several planned Siri features in 2024, including personal context awareness and the ability to understand what appears on a user’s iPhone or iPad screen.
The same report said Siri may also receive a dedicated app built to compete with ChatGPT, Claude, and other AI assistants. Yahoo Tech added that Apple plans to place Siri inside the iPhone’s Dynamic Island and let users swipe down from the middle of the screen to begin a conversation.
Nvidia gains another high-profile AI customer
The Information’s report also adds weight to a 2025 rumor that Apple had ordered 250 Nvidia NVL72 servers at about $4 million each. Nvidia introduced the Blackwell B200 architecture in 2024 for large language model training and inference.
Apple has not confirmed the reported arrangement. Still, the WWDC keynote on June 8 is expected to show how far the company has moved in rebuilding Siri after nearly two years of delays.
