Apple Partners with Google and Nvidia to Develop Cutting-Edge AI Cloud Models
0xBroomberg
Apple confirmed at WWDC that its most powerful cloud AI model runs on Nvidia GPUs inside Google Cloud — the first official admission that Apple Intelligence depends on rivals' infrastructure, raising a core question: can Apple's privacy promises hold on someone else's hardware?
What exactly did Apple announce?
Apple's most advanced cloud model is called AFM Cloud Pro. It runs on Nvidia GPU clusters hosted in Google Cloud.
This is Apple's first official confirmation that Nvidia chips power part of the Apple Intelligence stack.
This means → Apple's AI ceiling is not set by its own silicon — the strongest model needs Nvidia GPUs.
Is Apple just using Google's Gemini?
No. Senior VP Craig Federighi clarified: Apple Intelligence runs Apple's own models, not a public version of Gemini.
Google supplies the underlying cloud compute platform, not a ready-made AI model — Apple runs its own building on Google's foundation.
In plain terms = the market expected Apple to call Gemini directly. The actual partnership is far deeper and less visible: Google provides hardware and cloud; Apple provides the model and the privacy layer.
How is the privacy problem solved?
Apple required Nvidia chips to be configured so they cannot read user content on the server.
Nvidia developed "ambiguous confidential compute" — a technique that lets a chip process data without ever seeing the data's contents — specifically to meet this requirement.
Apple's OS includes a built-in "system orchestrator" that routes each AI query to on-device or cloud models based on compute needs and the volume of personal data involved.
This means → privacy protection rests not on avoiding third-party hardware, but on layers of architectural encryption — yet whether this promise truly holds is something no independent party can currently verify.
What does this mean for each company?
Nvidia: Apple joining the client list further cements Nvidia GPUs as irreplaceable in cloud AI training and inference.
Google: the model isn't Google's, but landing Apple as a cloud customer directly boosts Google Cloud revenue expectations.
Apple: it frames "privacy" as its counter-narrative to OpenAI and Anthropic's "AI for AI's sake" approach — but the fact that its strongest model runs on rivals' infrastructure dents the credibility of that narrative.
What follow-up signals should the market watch?
Apple's privacy promise rests on Nvidia's "ambiguous confidential compute" — a new technology that has not yet undergone large-scale independent audit.
This reflects an emerging pattern in AI: the company that talks most about privacy turns out to be the most dependent on third-party infrastructure.
In plain terms = Apple's story is "we care about your privacy more than anyone" — but the hardware and cloud behind that story belong to someone else. That tension won't fade; it will only grow as more user data moves to the cloud.
Content is for reference only, not financial advice.