An Apple research study claims that the company trained its most recent AI models using Google’s tensor processing units. Some industry observers were taken aback by the decision, given that NVIDIA commands a sizable market share of over 80% of the chip market and is widely recognized for providing the most cutting-edge processors.
Apple hasn’t acknowledged explicitly that they’ve employed Google chips, but there was no mention of NVIDIA technology in the study that Apple released. Apple said in the article that 8,192 TPUv4 processors were utilized for its server AI model and 2,048 TPUv5p chips were used to create an AI model for iPhones.
Since NVIDIA sells chips independently without requiring users to use any particular platform, Google only provides clients with direct access to their TPUs through its Google Cloud Platform, where applications must be developed. Early in the year, Reuters revealed that Apple was using Google chips; however, the full scope of the supply was unknown until the publication of the Apple study document.
A large portion of Apple’s AI research is directed at improving the AI services that the company provides to customers, as evidenced by the June unveiling of the Apple Intelligence system.
The potential that Apple Intelligence holds for our users is immense, and it’s exciting to watch our developers start creating with it. Tim Cook, the CEO of Apple, stated in a post on X that “we are excited to see the amazing things they create.”
Although the full scope of AI services’ integration into the Apple ecosystem is still unknown, at this point Apple anticipates that AI-powered solutions will assist customers in performing a variety of daily tasks and making use of new capabilities.