On May 28, 2026, MacRumors cited The Information reporting that Apple plans to place on-device AI capabilities at the core of WWDC 2026 in June: emphasizing the privacy and cost advantages of native inference with iPhone, Apple Watch, and Mac chips, while continuing to delegate complex requests to the cloud.
The key to this message is not just "Apple is talking about AI," but that Apple's external narrative may shift from chasing chatbots to emphasizing devices, chips, and privacy architectures. For users, this is more important than simply releasing a more chatty Siri, because on-device AI determines which requests can be faster, more private, and less dependent on data centers.
Gemini doesn't necessarily stand in the spotlight
The report mentions that Apple may use Google's large Gemini model to train smaller local models, meaning that by distilling the model's capabilities, large models are compressed to the range where Apple devices can operate. From this perspective, Gemini's role may not be a direct chat entry point for users, but more like a source for underlying training and capability transfer.
This also explains why Apple continues to emphasize its hardware advantages. Complete large models often require massive computing power, while local models must balance latency, power consumption, privacy, and capability. If Apple could perform frequently used tasks locally, it could turn "AI not necessarily in the cloud" into a differentiated selling point.
Cloud solutions are also evolving
The report also states that some cloud requests may use Google Cloud and introduce Nvidia's confidential computing technology to encrypt data and models being processed. Compared to Apple's earlier emphasis on Private Cloud Compute, this means the underlying implementation may be more mixed: the brand still emphasizes privacy, but the infrastructure doesn't necessarily depend solely on Apple.
This signals clearly to the industry: on-device AI and cloud AI are not a binary choice. In the future, phones, computers, and watches will handle some high-frequency, low-risk tasks, while the cloud will handle more complex reasoning and generation. The real competitive point becomes how the system determines where tasks should run and whether users can understand where the data has gone.
Why it deserves attention
Since Apple Intelligence's release in 2024, the delay in Siri's personalization capabilities has been affecting market confidence. If WWDC 2026 chooses to reshape the narrative with on-device AI, Apple is not competing head-to-head with ChatGPT and Gemini for chat boxes, but rather putting AI back into the device ecosystem. For developers and regular users, the next thing to watch is not how flashy the demo is, but which features can run offline or semi-offline, which requests go to the cloud, and whether permission instructions are clear enough.
Source of information
Source: MacRumors report, May 28.
