Apple has made a strategic pivot in its AI software infrastructure by selecting Google's tensor processing units (TPUs) over industry leader Nvidia's GPUs. This decision marks a significant shift in the AI processor market, where Nvidia currently holds an 80% share.
According to a research paper published on Monday, Apple is deploying two types of Google's TPUs: 2,048 TPUv5p chips for devices like iPhones and 8,192 TPUv4 processors for server-side AI models. Unlike Nvidia, which specializes in graphics processing units, Google's TPUs are accessible through its Google Cloud Platform, necessitating the use of Google's cloud services to leverage the hardware.
The full scope of Apple's reliance on Google hardware remained undisclosed until the recent publication. Both Google and Nvidia have declined to comment on Apple's choice.
Apple's engineers highlighted that Google's chips have the potential to support the development of even more advanced AI models in the future, suggesting a long-term collaboration aimed at enhancing AI capabilities across Apple's product line.
Reference(s):
Apple picks Google's chips over Nvidia for AI software infrastructure
cgtn.com
