Apple Inc. has chosen to use chips designed by Google rather than the industry leader Nvidia to build crucial components of its artificial intelligence (AI) software infrastructure. This decision, detailed in an Apple research paper published on Monday, marks a shift in the tech giant’s approach to developing its forthcoming suite of AI tools and features.
Apple’s reliance on Google’s cloud infrastructure stands out, given Nvidia’s dominance in the AI processor market. Nvidia’s chips are highly coveted for AI applications, holding approximately 80% of the market, including those from Google, Amazon, and other cloud computing companies.
Interestingly, Apple’s research paper did not explicitly state that it completely bypassed Nvidia chips. However, the description of the hardware and software infrastructure supporting its AI tools notably lacked any mention of Nvidia’s hardware.
According to the iPhone maker, it employed two variants of Google’s tensor processing units (TPUs) organized in extensive chip clusters to train its AI models. For the AI model designed to run on iPhones and other devices, Apple utilized 2,048 TPUv5p chips. Meanwhile, for its server AI model, the company deployed 8,192 TPUv4 processors. Nvidia, on the other hand, does not manufacture TPUs but focuses on graphics processing units (GPUs), which are extensively used in AI tasks.
Unlike Nvidia, which sells its chips and systems as standalone products, Google offers access to its TPUs through the Google Cloud Platform. Customers must build their software on Google’s cloud platform to utilize these chips, presenting a different business model compared to Nvidia’s.
Apple’s choice to lean on Google’s technology over Nvidia’s for its AI infrastructure could signal a shift in the competitive landscape of AI hardware, as it highlights the evolving strategies of leading tech companies in their quest for AI supremacy.