TLDR
- Apple disclosed that AFM Cloud Pro, its cutting-edge AI model, operates on Nvidia GPUs accessed via Google Cloud infrastructure
- Nvidia (NVDA) stock climbed approximately 0.9% to $210.44 during premarket hours following the announcement
- The tech giant isn’t purchasing chips directly from Nvidia—instead using Google’s cloud platform to access the hardware
- This marks Apple’s inaugural AI model officially powered by chips not manufactured by Apple itself
- Nvidia’s advanced “ambiguous confidential compute” capability enabled the partnership by satisfying Apple’s rigorous privacy requirements
Apple has publicly acknowledged that its flagship AI model operates on Nvidia hardware — yet the market’s response has been relatively muted, and the reasoning is clear.
Nvidia (NVDA) stock registered a gain of roughly 0.9% to $210.44 during Tuesday’s premarket session after the disclosure emerged at Apple’s WWDC conference on Monday.
The AI system is known as Apple Foundation Model Cloud Pro, or AFM Cloud Pro for short. According to Apple AI executive Amar Subramanya, it rivals the capabilities of Google’s frontier Gemini models.
The model operates in cloud environments using Nvidia GPUs — but there’s an important qualifier. Apple isn’t purchasing these processors directly from Nvidia. Rather, the company accesses them via Google Cloud as part of its Private Cloud Compute framework.
This indirect arrangement probably accounts for the tepid market response. The collaboration might not generate substantial fresh chip revenue for Nvidia, at least not in the immediate term.
Sebastian Marineau-Mes, Apple’s VP of software, explained that Apple expressly wanted Nvidia’s newest processors but required them deployed in a more secure configuration — one preventing the chips from accessing server data.
How Nvidia’s Privacy Innovation Enabled the Partnership
A breakthrough Nvidia technology known as “ambiguous confidential compute” provided Apple and Google with the technical infrastructure necessary to create a system aligned with Apple’s stringent privacy protocols.
“We wanted to avail ourselves of the latest technology from Nvidia, and so we set out to extend private cloud compute to third-party cloud,” Marineau-Mes said.
Craig Federighi, Apple’s software SVP, emphasized that Apple Intelligence relies on Apple’s proprietary models — not Google’s ready-made Gemini products, contrary to widespread speculation. Google’s infrastructure was utilized to assist in training Apple’s custom models, not to substitute them.
“These four models… are custom built for Apple Silicon, trained using proprietary data with reinforcement learning and refined using outputs from Gemini frontier models,” Subramanya explained.
Apple and Google had previously unveiled a comprehensive partnership in January. Monday’s WWDC presentation marked the initial public confirmation of Nvidia’s participation.
Apple’s Minimalist AI Approach
Apple has intentionally steered clear of the billions-dollar AI infrastructure investments that competitors have embraced. Federighi delivered a veiled criticism of that strategy during his keynote address.
“Some appear to be racing forward, seemingly pursuing AI for the sake of AI, without clear regard for the people — all of us — that it’s ultimately meant to serve,” he said.
Apple’s strategy centers on privacy-centric AI: systems that function locally whenever feasible, accessing cloud resources only when necessary, while utilizing personal information like calendar appointments and text messages to customize responses.
For Nvidia, securing Apple as a client — even through an intermediary — represents a significant achievement. It reinforces Nvidia’s position as the premier hardware provider for AI applications, despite mounting competition from Intel, AMD, and custom chip developers.
Nvidia stock concluded the trading day up 1.73%. Apple shares declined 1.89%.
