Apple’s revamped Siri to run on Google Cloud with Nvidia Blackwell chips

Tech giant Apple is furthering plans to power its Siri assistant revamp using Nvidia’s Blackwell B200 processors hosted in Google’s data centers, which points to a U-turn from the company’s regular strategy of having control over its entire tech stack.

The announcement is expected to come alongside a preview of iOS 27 and the initial views of a redesigned and overhauled Siri assistant after an almost 2 year delay, at Apple’s Worldwide Developers Conference keynote on June 8.

Apple turns to outside infrastructure, raising privacy concerns

During Apple Intelligence’s unveiling at WWDC 2024, senior vice president Craig Federighi stressed that cloud-based AI queries would run exclusively on Apple’s own servers through a system called Private Cloud Compute (PCC).

This system, built on Apple’s Mac-series silicon, was designed to process prompts without keeping user data or allowing it get into the hands of third parties.

Two years later, things appear to be taking a different route. The tech giant attempted to run a modified version of Google’s Gemini model on its Private Cloud Compute infrastructure but found the performance too slow to be practical, according to The Information’s reporting. The company will instead direct cloud-based Siri queries through Google’s collection of Nvidia Blackwell B200 data center chips.

To address the obvious privacy concern, the tech firm will enable Nvidia’s hardware-based confidential compute feature on the chips the company plans to use. The confidential compute technology encrypts user data as it is processed on cloud, a safeguard meant to further guarantee privacy beyond only Apple’s hardware. It remains unclear how this new flow would work with the existing Private Cloud Compute branding, however, MacRumors reported that there is a high chance Apple keeps the PCC name regardless of the change.

Nvidia comes out on top

This new arrangement stresses the scale of Nvidia’s dominance in AI infrastructure. It represents a lot if even Apple, a company with $3 trillion in market capitalization and its own custom chips and processors design team, needs Blackwell chips to run competitive LLMs at scale.

The B200 processors, which succeeded Nvidia‘s Hopper architecture in 2024, were built specifically to accelerate and improve LLM training and inference.

A 2025 rumor that Apple was purchasing 250 Nvidia NVL72 servers at roughly $4 million each is now beginning to look very credible in light of this report.

Revamped Siri’s features

The upgrades and overhaul planned for the Siri assistant represent the biggest change to Siri since its launch in 2011. A lot of features were first promised in 2024, including the ability to understand personal context and analyze on-screen content on iPhones and iPads, and these are finally expected to arrive.

Siri is also expected to get a dedicated app designed to compete directly with ChatGPT, Claude, and other popular generative AI assistants. The assistant will also move into the iPhone’s Dynamic Island, and users will gain the ability to swipe down from the center of the screen to start a conversation, Yahoo Tech reported.

Apple is targeting a September release for the updated Siri, which aligns with the expected launch of iOS 27. The WWDC 2026 keynote begins at 10:00 a.m. Pacific on Monday, June 8. The Siri overhaul is expected to be a centerpiece of the presentation alongside iOS 27, macOS updates, and even more Apple Intelligence improvements.