Report: Apple Plans to Make On-Device AI a Key WWDC Focus

Source: MacRumors

Overview

Apple is expected to use next month’s Worldwide Developers Conference (WWDC) to showcase its on‑device AI capabilities, positioning them as a competitive advantage built on 15 years of custom silicon expertise. The company aims to demonstrate that AI models can run locally on iPhones, Apple Watches, and Macs, offering privacy‑preserving and cost‑saving benefits compared with cloud‑based processing.

On‑Device AI Strategy

• Local inference: Apple plans to run many AI queries directly on its devices, reducing reliance on data‑center infrastructure.
• Cloud fallback: Complex queries will still be handled in the cloud, but the emphasis will be on keeping most processing on‑device.
• Privacy focus: Running models locally is presented as a way to protect user data while still delivering AI features.

Partnership with Google

• Model distillation: Under an agreement with Google, Apple will use a large version of Google’s Gemini model to train a smaller, distilled version that can operate on Apple hardware.
• Acquisition scouting: Apple is reportedly evaluating companies such as Liquid AI, a Massachusetts startup that specializes in on‑device AI, to accelerate its model‑shrinking efforts.
• Nvidia compute in Google Cloud: Apple has approved the use of Nvidia’s confidential compute technology within Google Cloud to process the larger Gemini‑based model. This adds encryption for data and model security, with a modest performance cost.

Challenges and Limitations

• Scale of Gemini: The full Gemini model contains trillions of parameters, and Apple has struggled to run it on its Private Cloud Compute infrastructure, which uses the same Apple silicon chips found in Macs.
• Shift from original plan: Apple’s initial “Apple Intelligence” announcement promised that all cloud‑bound queries would be handled exclusively by its Private Cloud Compute running on Apple silicon. The new partnership suggests a pivot, though Apple may retain the Private Cloud Compute branding.

Future Outlook

• WWDC 2026: Scheduled for June 8, the event is expected to reframe Apple’s AI narrative, reintroduce delayed features, and debut new on‑device capabilities.
• Previous rollout: Apple Intelligence was first announced at WWDC 2024, but its rollout has been slowed by a lukewarm response to early features and delays to the more personal version of Siri.

Apple’s focus on on‑device AI aims to differentiate its ecosystem by leveraging custom silicon, privacy advantages, and strategic collaborations with Google and Nvidia.