Apple Set to Pitch On Device AI as Its Edge at WWDC 2026

Apple plans to use next month's Worldwide Developers Conference to frame its ability to run AI directly on its devices as a competitive advantage

Marcus Reed·May 28, 2026·3 min read

Apple plans to use next month's WWDC to frame its ability to run AI directly on its devices as a competitive advantage, according to a new report from The Information.

People familiar with the company's plans say Apple will lean on roughly 15 years of custom silicon work to argue that processing AI queries locally is a better path than the cloud. Cloud processing will still handle the most complex requests, but Apple is expected to present local inference as a more private and less costly alternative to the large data center buildouts its rivals have pursued.

Training a smaller model

To make that case, Apple is reportedly turning to Google. As part of an agreement reported earlier this year, the company is said to be using a large version of Google's Gemini model to train a smaller, distilled version that can run on Apple hardware.

Apple is also said to be looking at acquisitions to speed up that work. One company it has reportedly considered is Liquid AI, a Massachusetts startup focused on running AI models locally on devices.

When the cloud is still needed

Not every request can be handled on an iPhone or a Mac. For the larger model based on Gemini, Apple is believed to have approved the use of Nvidia's confidential compute technology inside Google Cloud.

The feature encrypts data and AI models while they are being processed. According to the report, that adds a modest performance cost but offers stronger privacy protections.

This marks a notable shift from Apple's original Apple Intelligence plan. When the system was announced, Apple said cloud bound queries would be handled only by its own Private Cloud Compute infrastructure running on Apple silicon. People familiar with the partnership told The Information that Apple is likely to keep the Private Cloud Compute name despite the change.

The limits of running locally

There are said to be real limits to how far Apple can take on device processing. Google's full Gemini model runs into the trillions of parameters, and The Information reports that Apple has struggled to run it on its Private Cloud Compute hardware, which uses the same Apple silicon chips found in Mac computers.

Reframing Apple Intelligence

Apple Intelligence was first announced at WWDC 2024, but the rollout has been slowed by a lukewarm response to its early features and a lengthy delay to the more personal version of Siri.

WWDC 2026 runs from June 8 to June 12. Apple is now expected to use the event to reset the story, reintroduce the delayed features, and debut new ones. Apple has not commented on the report.

My Favorite iPhone Accessories:

My Favorite Mac Accessories:

My Favorite iPad Accessories:

Note: That Apple Guide is an affiliate partner with some of these vendors. When you click a link and make a purchase, we may receive a small payment, which helps us keep the site running. At no extra cost to you.