Apple Pivots to Google for AI
Apple is reportedly shifting from its own cloud infrastructure to using Google's servers for its next-generation Siri and AI features. The move comes as Apple is also set to replace Core ML with a new "Core AI" framework in iOS 27, which will likely be powered by Google's Gemini models, signaling a major strategy shift in the consumer AI race.
Apple's pivot to Google's AI infrastructure follows a period of significant internal challenges. The company's AI/ML group has been described as "AIMLess" by some employees, with frustrations over a lack of clear direction and multiple technical pivots on Siri's development. This reportedly led to project delays and prompted some staff to depart. The decision to partner with Google comes after Apple's own large language model, codenamed "Ajax," reportedly struggled to match the performance of models from competitors like OpenAI. While Apple has been developing on-device AI for years, its strict privacy-first stance has limited the data available for training large-scale, cloud-based models, a domain where Google has a distinct advantage. The multi-year deal involves licensing a custom version of Google's Gemini models to power Apple Intelligence and the long-awaited Siri overhaul. This partnership is not Apple's first foray into third-party AI; the company already has a deal with OpenAI to integrate ChatGPT into its ecosystem. However, the deeper integration with Google's foundation models for core features represents a more significant strategic shift. Apple will maintain control over its Private Cloud Compute infrastructure, ensuring user data remains private and is not accessible to Google. Concurrently, Apple is doubling down on its own hardware with "Project ACDC" (Apple Chips in Data Centers), an initiative to develop custom chips for AI processing in its servers. This move, similar to Google's own TPU development, aims to vertically integrate Apple's AI stack, from on-device Apple Silicon to the data center, reducing reliance on third-party hardware for AI inference tasks. For developers, the transition from Core ML to a new "Core AI" framework is expected to be announced at an upcoming Worldwide Developers Conference. This rebranding signals a broader focus beyond traditional machine learning and is intended to make it easier for developers to incorporate advanced AI models, including third-party ones, into their apps for platforms like iOS 27.
