Apple May Use Google Cloud for AI Siri

This potential partnership extends beyond just licensing Google's Gemini AI models; Apple is reportedly in talks with Google to host the next-generation Siri on Google's own servers. To maintain Apple's stringent privacy standards, the discussions include setting up dedicated servers for Siri within Google's data centers. Historically, Apple has utilized a hybrid cloud approach, relying on both its own data centers and third-party providers like Amazon Web Services (AWS) and Google Cloud for services such as iCloud. This move to potentially run a core service like Siri on a competitor's infrastructure highlights the immense computational power required for advanced AI. The discussions come as Apple's own AI infrastructure, known as Private Cloud Compute, has reportedly faced challenges, including low utilization and technical constraints. Some reports suggest that Apple's AI infrastructure is "beginning to decay," prompting a reevaluation of its cloud strategy as it decommissions older hardware. Google's advantage lies in its specialized hardware, particularly its Tensor Processing Units (TPUs), which are custom-built to efficiently run large language models like Gemini. This could provide the necessary scalability and performance for a revamped, more conversational Siri, a project that has reportedly faced delays. At its 2024 Worldwide Developers Conference, Apple introduced "Apple Intelligence," a system blending on-device processing for simpler tasks with its Private Cloud Compute for more complex queries. The potential integration with Google's powerful infrastructure would be reserved for the most demanding generative AI features. This is not Apple's first foray into using external AI models, having already announced plans to integrate Google's Gemini into its products. However, hosting the core of Siri on Google's servers would represent a much deeper, strategic dependence on a key rival.