Apple will reportedly make on-device artificial intelligence the centerpiece of its upcoming Worldwide Developers Conference, leveraging decade-and-a-half of custom silicon development to argue that local AI processing offers privacy and cost advantages over cloud-based competitors. The company plans to demonstrate how chips built for iPhones, Apple Watches, and Macs can handle AI queries directly on devices rather than routing them through distant servers.

According to sources familiar with the plans, Apple intends to use Google's Gemini model to create smaller, distilled versions optimized for local execution while pursuing acquisitions to advance this model-shrinking effort. One target reportedly includes Liquid AI, a Massachusetts startup specializing in device-local AI. For queries requiring cloud processing, Apple has reportedly approved Nvidia's confidential compute technology within Google Cloud to encrypt data during processing, though this adds performance overhead.

This represents a shift from Apple's original Apple Intelligence announcement, which promised exclusive use of Private Cloud Compute infrastructure for all cloud-bound requests. Sources indicate Apple has struggled to run Google's massive Gemini model—spanning trillions of parameters—on its own Private Cloud Compute systems, which rely on standard Mac silicon. The company appears likely to maintain the Private Cloud Compute branding despite partnering with Google Cloud.

Apple Intelligence debuted at WWDC 2024 but faced lukewarm reception and delays to personalized Siri features. With WWDC 2026 beginning June 8, Apple aims to reset expectations, revive postponed capabilities, and introduce new on-device AI functionality that showcases its silicon advantage.