Apple revealed in a technical paper on Monday that it used Google’s custom-designed chips to train two of its artificial intelligence models. The decision highlights the growing competition in the AI infrastructure space as companies seek diverse and efficient solutions. The paper detailed that Apple’s AI models were trained on Cloud TPU clusters.
These clusters allowed Apple to train the models efficiently and at scale. The Apple Foundation Model (AFM) for on-device use was trained on a single “slice” of 2048 TPU v5p chips. The AFM server model was trained on 8192 TPU v4 chips configured to work together as eight slices through a data center network.
TPUs, first introduced by Google in 2015, have matured into one of the leading custom chips for AI alongside Nvidia’s GPUs. Despite developing its own TPUs, Google remains a significant customer of Nvidia, using both companies’ technologies in its cloud services.
Google’s TPUs bolster Apple’s AI
Apple has been slower than its peers to publicly embrace generative AI. However, on Monday, the company introduced several new AI features, such as an updated interface for Siri, improved natural language processing, and AI-generated text summaries. The upcoming year will see more functions based on generative AI, including image and emoji generation, and a more powerful Siri capable of performing personal tasks within apps.
In their ongoing development, Apple is employing TPUs for AI training while inferencing will partially occur on Apple’s own data center chips. Apple published an initial paper about its AI system in June and has now provided deeper insights into its use of TPUs. The company is aligning its strategy with the evolving AI landscape to ensure it can compete in delivering cutting-edge AI technologies.
Apple is scheduled to report its quarterly earnings after the close of trading on Thursday, which will add context to its latest advances and strategic moves in the AI sector. The collaboration between Apple and Google marks a rare instance of two tech giants working closely together in a highly competitive sector. Industry experts speculate that the partnership could have significant implications for the development of future AI functionalities in Apple’s ecosystem, potentially enhancing features in various Apple products, from iPhones to Macbooks.
