Apple’s recent research paper revealed that the company is not using Nvidia GPUs to train its AI models, opting instead for chips designed by Alphabet’s Google. This marks a significant shift given Nvidia’s dominance in the AI processor market.
In the paper published on Monday, Apple detailed its use of Google’s tensor processing units (TPUs) to build its AI software infrastructure. Specifically, Apple utilized 2,048 TPUv5p chips for AI models on devices like iPhones, and 8,192 TPUv4 processors for its server AI model. The decision to employ Google’s hardware highlights a departure from Nvidia’s GPUs, which are widely regarded as the leading processors for AI applications. However, Apple did not explicitly state the absence of Nvidia chips, merely omitting any mention of them in its hardware and software descriptions.
Google’s TPUs, available through its cloud platform, differ from Nvidia’s standalone GPU products. Customers using Google’s TPUs must build software within the Google Cloud Platform to leverage the hardware. Despite this reliance on Google’s infrastructure, neither Google nor Nvidia provided comments on the matter.
Apple to Launch New AI Suite to Beta Users This Week
Apple’s engineers noted that the potential for creating even larger and more advanced models exists with Google’s TPUs. This disclosure follows Apple’s unveiling of new AI features at its June developer conference, including the integration of OpenAI’s ChatGPT technology into its software.
Apple is set to introduce parts of its AI suite, Apple Intelligence, to beta users this week. The research paper elaborates on the company’s strategic choice to use Google’s TPUs, a decision initially reported by Reuters in June but fully detailed only now.
Featured Image courtesy of Bloomberg
Follow us for more updates on Apple.