Apple targets AI marketplace with Siri extensions and on‑device models

Apple is reshaping its AI strategy by turning Siri into a gateway for a new AI marketplace, where developers can plug models like ChatGPT, Gemini or Claude directly into the assistant. The approach lets Apple earn commissions on subscriptions without building its own frontier AI model, and it leans on the company’s strength in on‑device processing [1].

Key takeaways

• Siri Extensions will enable users to load external AI models and switch between them within Siri [1].
• Apple’s AI engine will run lightweight versions of Google’s Gemini locally on iPhone, Mac and Apple Watch [2].
• The on‑device focus supports Apple’s “data stays on the device” privacy promise while reducing the need for large cloud data centers [1][2].
• More complex tasks will still be offloaded to Google Cloud, creating a hybrid processing model [2].
• Apple acknowledges its internal AI capabilities lag behind rivals, prompting a pragmatic platform‑first strategy [1].

Siri as an AI Marketplace

Apple’s plan treats the App Store model as a template for AI services, positioning Siri as the entry point for a marketplace of conversational agents. Developers can package models as “Siri Extensions,” allowing iPhone users to install and activate services such as ChatGPT, Gemini or Claude without leaving the Siri interface. Apple would collect a commission on each subscription, mirroring the App Store’s revenue‑sharing structure [1]. This strategy sidesteps the need for Apple to develop a proprietary frontier model, a task it admits is challenging given the scale of infrastructure required [1].

On‑Device AI Engine

At WWDC, Apple is expected to showcase a new AI architecture that runs primarily on the device. A lightweight variant of Google’s Gemini model has been optimized for Apple’s neural processing units, enabling common AI tasks to be performed locally on iPhone, Mac and Apple Watch [2]. The company emphasizes that this keeps user data on the device, aligning with its long‑standing privacy stance. For more demanding computations, the system will securely forward data to Google Cloud, creating a two‑layered pipeline that balances privacy with performance [2]. Apple’s silicon, designed for efficient on‑device inference, is highlighted as a core advantage in this approach [1].

Why it matters

By converting Siri into a platform for third‑party AI services, Apple can monetize AI without the heavy R&D burden of building its own large‑scale models. The on‑device focus reinforces Apple’s privacy narrative and may appeal to users wary of cloud‑based AI. However, the reliance on external models and a hybrid processing model means Apple’s AI capabilities will be closely tied to partners like Google, at least in the near term. Future developments will likely hinge on how well the marketplace attracts developers and how effectively the on‑device models meet user expectations for speed and accuracy.