Apple's AI reveal hands an unexpected win to Google, NVIDIA
Apple Inc. revealed that its advanced AI model AFM 3 Cloud Pro will run on NVIDIA chips via Google Cloud, signaling a shift from its private cloud strategy. Analysts are divided on the implications for margins and monetization, though JPMorgan maintains a positive long-term outlook. Siri AI features will enter beta later this year, excluding the EU and China initially.

*this image is generated using AI for illustrative purposes only.
Apple Inc. unveiled its biggest Siri AI upgrade in years at the Worldwide Developers Conference, but the more consequential disclosure came behind the scenes: the company's most advanced AI model will run on NVIDIA Corp. chips through Alphabet Inc.'s Google Cloud rather than on its own silicon. This shift marks a significant departure from Apple's longstanding hardware strategy, raising questions about future capital spending and the company's ability to monetize AI through its services business. Apple shares were down 0.12% at $290.20 during premarket trading on Wednesday, while NVIDIA shares rose 0.77% to $210.24 previously. Apple fell 3.64% in the previous session, underperforming the Nasdaq and landing at the bottom of the Magnificent Seven.
Siri AI capabilities and limitations
CNBC's MacKenzie Sigalos noted that Apple unveiled functional generative AI branded as Siri AI at WWDC. The new Siri is more conversational, features a standalone app, can understand what appears on a user's screen, and can draw on personal context. However, Apple did not demonstrate a fully agentic assistant capable of tasks such as booking dinner reservations or calling a car. The lack of groundbreaking agentic capabilities and the absence of a clear path to monetize AI through Apple's services business influenced the stock reaction. Siri AI features are available for developer testing, with a beta release later this year for users with supported devices set to English.
Infrastructure and partnership
Apple announced a new generation of its own large language models built with help from Google Gemini, which will run on-device and through private cloud compute. However, Apple disclosed that its most advanced frontier AI model, AFM 3 Cloud Pro, is designed to run on NVIDIA graphics processors rather than its own silicon. Apple waited until a post-keynote briefing to discuss AFM Cloud Pro, a model it says is on par with frontier large language models and powerful enough that Apple will need NVIDIA chips and Google Cloud for some compute. Sigalos said Alphabet emerged as a WWDC winner because Apple is relying on Google Cloud and Google's models after acknowledging that its private cloud cannot support heavier AI workloads.
Analyst perspectives on strategy and margins
JPMorgan senior analyst Samik Chatterjee said Apple's Siri overhaul remains a long-term game changer, even though the stock fell after the WWDC reveal. JPMorgan maintained an Overweight rating on Apple with a $325 price forecast. Gabelli Funds' John Belton said Apple needs to show whether its revamped Siri is becoming a true platform or remains just a feature, noting that Apple needs to catch up in the age of agentic AI. Bernstein's Mark Newman said Apple's partnership approach is the right strategy because it gives the company access to advanced models such as Google's while avoiding the massive capital spending of hyperscalers. He said Apple spends about $14 billion a year on capital expenditures, far below the $200 billion being spent by some competitors.
Services monetization and availability
Sigalos said investors are focused on Apple's nearly 50% gross margin and are asking what happens if more advanced Siri and image-generation workflows increasingly run through Google Cloud systems and NVIDIA chips. She said inference is cheaper than training, but still creates a different cost structure than Apple's original on-device and private-cloud AI pitch. Apple has not launched a separate Apple AI subscription, though it is using iCloud Plus to test a paywall for heavier AI usage. Compatibility requires iPhone 16 models or later, iPhone 15 Pro, iPhone 15 Pro Max, and specific iPad and Mac models with M1 or later chips. Apple noted that Siri AI will not be available initially in the EU for iOS and iPadOS, nor in China, due to regulatory requirements.
Apple Active Device Base Growth
| Year | Annual Increase |
|---|---|
| 2022 | 9% |
| 2023 | 11% |
| 2024 | 10% |
| 2025 | 7% |
| 2026 | 6% |
How will the reliance on NVIDIA chips and Google Cloud impact Apple's profit margins as AI usage scales?
Will Apple eventually introduce a dedicated subscription tier for advanced AI features to offset cloud infrastructure costs?
What are the long-term strategic risks of Apple depending on competitors like Google for its most advanced AI workloads?























