Apple’s big yearly software event looks set to focus a lot on on-device AI, with reports suggesting the company’s in-house chip architecture will give it a key advantage over rivals.
WWDC 2026 kicks off on 9 June, with iOS 27 for iPhone and software updates for iPad, Mac, Apple Watch, and Apple Vision Pro all expected to be previewed at the event, and AI is widely anticipated to be integrated into pretty much every new feature Apple introduces.
The shift toward on-device processing addresses a core limitation of most AI implementations, where queries travel to remote data centres and back before a result reaches the user, introducing latency that depends entirely on network quality and connection speed.
Apple’s silicon lineup, which powers everything from the iPhone through to the Mac, carries enough processing headroom to handle AI inference locally, cutting out that round trip entirely and keeping the workload on the device itself.
Privacy and cost implications
That architectural difference carries particular weight on the privacy front, since on-device processing means user data never leaves the hardware, removing the exposure point that cloud-based AI systems create when queries pass through third-party infrastructure.
Running AI on-device also removes the per-query cost of data centre processing, a significant consideration as Apple scales Apple Intelligence features across hundreds of millions of active devices worldwide.
According to The Information, Apple is working with a version of Google’s Gemini model as a reference point for training a smaller, locally-capable model, and the company is also reportedly evaluating acquisitions of firms with expertise in models optimised for on-device deployment.
Whatever Apple announces on 9 June, the features themselves will not reach users immediately; software updates across all platforms are expected to follow the standard testing cycle ahead of a likely September release.
The post WWDC 2026 is shaping up to be all about on-device AI appeared first on Trusted Reviews.