Google and Nvidia are helping Apple with Apple Foundation Model Cloud Pro, which Apple says is comparable to Gemini frontier models and runs on Nvidia GPUs
- Apple showed off Siri AI demos at its annual WWDC keynote in Cupertino, Calif. … Apple on Monday revealed what it's been working …
CNBCKif Leswing
Context & Ripple Effects
Apple’s AI plan has moved from a multiyear agreement to use Gemini and Google Cloud for a more personalized Siri toward deeper access to Gemini, including the ability to fine-tune it and create smaller distilled models. The related coverage consistently frames Google’s models as the foundation for Apple’s near-term Siri work.
The WWDC disclosure adds Nvidia as the compute layer for Apple Foundation Model Cloud Pro. It indicates that Apple’s model effort is being assembled through partnerships spanning a frontier-model provider and GPU infrastructure rather than relying on a single external service.
First-order effects
Apple can position Apple Foundation Model Cloud Pro as a cloud model tailored to its own products while benchmarking it against Gemini frontier models; Google and Nvidia become named contributors to that stack.
Nvidia GPUs become the stated hardware substrate for the service, tying Apple’s cloud-model rollout directly to Nvidia-based AI compute.
Second-order effects
Apple’s deeper ability to fine-tune and distill Gemini-derived capabilities gives it more control over how foundation-model technology is adapted for Siri and future Apple Intelligence features, even as Google remains a core supplier.
The arrangement reinforces the commercial value of pairing frontier models with Nvidia GPU capacity: Google supplies model capability while Nvidia supplies the compute layer needed to operate a comparable cloud offering.
Third-order effects
If this architecture persists, leading consumer-device companies may increasingly differentiate through proprietary model layers and product integration while sourcing core models and accelerator infrastructure from specialist partners.
That would make AI competition less a choice between fully in-house and fully outsourced systems, and more a contest over control of customization, deployment, and the user-facing assistant layer.
The trend: This is one data point in the shift toward vertically integrated AI products built atop shared frontier-model and GPU infrastructure.
Per Craig Federhigi on how much Google gemini stuff they use for Apple Intelligence: “we don't have the Gemini app as our app. In fact, none of that client code is part of how we run an iOS for these models. We use none of the models that Google deploys to their customers, nor
Proud to collaborate on the next frontier of private AI. 🔒 Apple is expanding Private Cloud Compute to @GoogleCloud using NVIDIA GPUs. https://security.apple.com/...
“AFM Cloud Pro seems to be based on Gemini foundation and data, but Apple did their own pre-training, post-training, RL, etc” Google shared its training data and training recipes DAYUM
Apple just clarified AFM Cloud is Apple's own model, trained with Gemini outputs AFM local models are entirely Apple models AFM Cloud Pro seems to be based on Gemini foundation and data, but Apple did their own pre-training, post-training, RL, etc
🔺NEW: Apple is expanding Private Cloud Compute (PCC) beyond our data centers. PCC on Google Cloud: NVIDIA Confidential Computing, Intel TDX, and Google's Titan chip, with capabilities that go far beyond a traditional confidential computing deployment. https://security.apple.com/.…
@BenBajarin Agree and I wonder if it is just a signal of how much NVIDIA has become synonymous with AI so you never go wrong in mentioning how you work with them
Apple Foundation Model Cloud Pro is the best Apple model, and runs on Nvidia GPUs in Google Cloud Apple Foundation Model Cloud and Cloud Image run on Apple Silicon. Both are private cloud compute.
AFM Core Advanced on-device model running on A19 Pro is a sparse model. It's 20B parameters. It's fully Apple designed. It is an MoE but when it processes the prompt, it only loads the parameters needed and locks them in. If it's 20B parameters total, but on a specific
@BenBajarin Apple distilled Google's models for its own use. Which is why it didn't talk much about Google today. I hear that Apple is still highly freaked out about the privacy implications of AI. And my consumer research shows that armies of “normies” are also freaked out, so t…
If you have followed Apple for any length of time you know they don't mention partner/providers lightly. Specifically calling out NVIDIA here as well is interesting IMO. We knew this, but to hear them call it out...
It's almost as if agentic AI workloads run best on Nvidia GPUs? CRAZY. ""For the most demanding tasks, including agentic tool-use and complex reasoning, we worked with Google and NVIDIA to extend our PCC infrastructure to Google Cloud systems using NVIDIA GPUs, while maintaining
WOW. Apple is now using... NVIDIA GPUs!?! Let's go. $AAPL $NVDA $INTC “Now, we are collaborating with Google and NVIDIA to run new Apple Intelligence workloads on Google Cloud, extending our industry-leading PCC privacy commitments to third-party data centers for the first tim…
Apple execs confirming using broader GCP and NVIDIA GPUs for Apple Private cloud compute: “This is our most capable model, with quality similar to Gemini Frontier models, and to bring this model to production, we work with both Google and Nvidia to extend our private cloud comput…