Apple's 'Core AI' to Replace Core ML

The strategic pivot to "Core AI" follows a major leadership shuffle. In December 2025, Apple announced that John Giannandrea, head of AI strategy since 2018, would retire in spring 2026. His role is being filled by former Google executive Amar Subramanya, who previously led engineering for the Gemini Assistant, signaling a move to accelerate product development. The transition addresses the growing limitations of the Core ML name and framework, first introduced in 2017. While Core ML excelled at on-device inference using the Apple Neural Engine, its "machine learning" branding became dated, and developers had to stitch together multiple APIs, including the newer Foundation Models framework, to build modern generative features. Underpinning the new framework is a multi-year partnership with Google, confirmed in a joint statement on January 12, 2026. This deal makes Google's Gemini models the technological base for the next generation of Apple Foundation Models, aiming to power a significantly more capable and personalized Siri. Despite the reliance on Google's core technology, Apple emphasizes that its privacy standards will be maintained. Apple Intelligence will continue to run on-device and through its Private Cloud Compute system, which uses Apple silicon servers, ensuring user data is not directly exposed to third-party cloud infrastructure. Apple's own on-device models are significantly smaller, around 3 billion parameters, to ensure they can run efficiently and privately on iPhones and iPads. The "Core AI" framework is designed to help developers more easily integrate these on-device models with larger, more powerful external models like Gemini. This technical overhaul is expected to debut at WWDC alongside iOS 27. Reports suggest iOS 27 will be a performance-focused release, similar to the "Snow Leopard" cycle, to ensure the speed and stability required to support more demanding, system-wide AI capabilities.