Apple is set to rely on Google’s fleet of Nvidia (NASDAQ: NVDA) chips to power its overhauled version of Siri when it launches in September, according to a report from The Information.
Queries requiring cloud-based processing will fall back on one of Google’s large Gemini models, under an existing agreement between the two companies.
Apple will tap into Google’s fleet of Nvidia Blackwell B200 data center chips, where user data will be encrypted using Nvidia’s hardware-based confidential compute feature.
Reportedly, Apple attempted to run a version of Google Gemini under Private Cloud Compute, but found it too slow to be usable in a consumer-facing product.
This development comes just days after Nvidia unveiled its most powerful AI model to date, Nemotron 3 Ultra, at Computex 2026 in Taipei on June 1.
CEO Jensen Huang presented the open-source model, which carries approximately 500 to 550 billion parameters, and is designed for advanced reasoning and complex agentic workflows.
Nemotron 3 Ultra sits at the top of a three-tier model family that also includes Nano and Super versions, giving developers and enterprises a range of deployment options.
Nvidia described the model as delivering up to 5x faster inference while reducing the cost of complex agentic tasks by as much as 30%.
The broader Nemotron 3 family surpassed 50 million downloads in the year leading up to April 2026, signaling strong traction among developers and enterprise clients worldwide.
Confidential compute is a security feature embedded inside Nvidia graphics processing units that encrypts data and AI models as they are actively being processed in the cloud.
When enabled, the feature slightly slows the processing of AI queries, but it could help Apple maintain its longstanding promises around user privacy in cloud environments.
Apple’s move has been described as a divergence from the company’s strategy of attempting to control all the critical ingredients to its products, raising questions about Private Cloud Compute’s future role.
A prior report claimed Apple was in the process of purchasing 250 Nvidia NVL72 servers, each costing around $4 million, lending further credibility to the emerging hardware relationship.
This scenario positions Nvidia as fundamental infrastructure for one of the largest consumer AI launches in recent memory, extending its role well beyond chip manufacturing.
With Apple leaning on Nvidia hardware for Siri and Nemotron 3 Ultra consolidating its software credibility, Nvidia now controls both ends of the AI stack at a pivotal moment for the consumer AI market.
Apple Intelligence was unveiled at WWDC 2024, but its rollout has been shadowed by a lukewarm reception and ongoing delays to the more personalized version of Siri that users were promised.
WWDC 2026 begins on June 8, when Apple is widely expected to revisit those delayed features and introduce new AI capabilities to reset the product narrative.
