Apple Inc. has used chips designed by Google instead of those by industry leader Nvidia for two key components of its artificial intelligence software infrastructure, reports revealed Tuesday, citing the iPhone maker’s latest research paper.
Despite the dominance of Nvidia, which controls about 80% of the market, along with chips from Google, Amazon (NASDAQ:), and other cloud computing companies, Apple chose Google’s cloud infrastructure.
According to Reuters and CNBC, the research paper did not explicitly mention that Nvidia chips were used, but it was noticeably absent of any mention of Nvidia hardware in its description of Apple’s AI tools and architecture. Instead, Apple relied on two types of Google’s tensor processing units (TPUs) to train its AI models, which were organized into large clusters of chips.
To develop the AI model that will be used on iPhones and other devices, Apple used 2048 TPUv5p chips. For the server-based AI model, Apple deployed 8192 TPUv4 processors.
Google’s latest TPUs are cost-effective, costing less than $2 per hour of use when reserved for three years in advance, according to Google’s website. The TPUs, which Google first introduced in 2015 for internal purposes and made available to the public in 2017, are now among the most advanced custom chips ever designed for AI.
By contrast, Nvidia focuses on designing graphics processing units (GPUs) that are widely used in AI applications. Unlike Nvidia, which sells its chips and systems as standalone products, Google offers access to its GPUs through Google Cloud Platform, which requires customers to build software within Google’s cloud ecosystem to take advantage of the chips.
In the research paper, Apple engineers noted that it would be possible to develop larger and more sophisticated models using Google chips than those discussed.
Apple is set to roll out parts of its Apple Intelligence suite to beta users this week.
The company announced several new AI features at its developer conference in June, including integrating OpenAI’s ChatGPT technology into its software.
Over the next year, the tech giant plans to launch AI-powered features like image and emoji generation, as well as an enhanced Siri that can access user information and perform actions within apps.