Apple has revealed that it relied on Google's Tensor Processing Units (TPUs) to train its new AI models, diverging from the industry norm of using Nvidia's GPUs. Apple used 2,048 TPUv5p chips for its on-device model and 8,192 TPUv4 chips for its server model. This move signals a potential shift in the AI chip market, as other tech giants like Microsoft, Meta, and Amazon are also exploring alternatives to Nvidia. Apple's decision to use Google's cloud infrastructure allowed for efficient and scalable training of its AI models, as detailed in a recent technical paper. The company plans to roll out features of its AI system, Apple Intelligence, to beta users this week.