Apple is set to emphasize its on-device AI capabilities at the upcoming Worldwide Developers Conference (WWDC), leveraging its 15 years of expertise in custom silicon to advocate for running AI models directly on devices rather than relying on the cloud. This approach aims to offer a privacy-focused and cost-effective alternative to the extensive data center infrastructures used by competitors.
Apple plans to demonstrate how its chips, designed for iPhones, Apple Watches, and Macs, provide an advantage in processing AI tasks locally. While cloud processing will still be necessary for complex queries, the company is working on a collaboration with Google to use a version of Google’s Gemini model to develop a smaller version suitable for local use on Apple devices. Additionally, Apple is exploring acquisitions to further its efforts in optimizing AI models for on-device use, with Massachusetts-based Liquid AI being a potential candidate.
Despite this focus on local processing, some tasks will still require cloud support. Apple has approved the use of Nvidia’s confidential compute technology within Google Cloud to process larger models, ensuring data and model encryption for enhanced privacy. This marks a shift from Apple’s initial Apple Intelligence announcement, where it promised that all cloud-bound queries would be managed by its own Private Cloud Compute infrastructure.
However, there are limits to Apple’s on-device processing capabilities. Running Google’s full Gemini model, which has trillions of parameters, poses challenges for Apple’s Private Cloud Compute infrastructure, which utilizes the same silicon chips as Mac computers.
Originally announced at WWDC 2024, Apple Intelligence faced delays and lukewarm reception. With WWDC 2026 commencing on June 8, Apple aims to reshape the narrative, reintroduce postponed features, and unveil new advancements.
