Apple used chips created by Alphabet’s Google rather than industry leader Nvidia to build its new artificial-intelligence software infrastructure, which will power its upcoming suite of AI tools and capabilities, according to an Apple research paper released on Monday.
Apple’s decision to employ Google’s cloud infrastructure is notable because Nvidia makes the most sought-after AI chips. Including chips made by Google, Amazon.com, and other cloud computing giants, Nvidia controls nearly 80% of the industry.
Apple did not expressly state in the research article that it does not employ Nvidia chips, but the description of the hardware and software infrastructure of its AI tools and features made no reference of Nvidia technology.
On Monday, Apple was unable to provide an immediate comment.
Apple stated in the study report that it trained its AI models using two flavours of Google’s tensor processing unit (TPU) that are organised in massive clusters of processors.
Apple utilised 2,048 TPUv5p chips to create the AI model, which will run on iPhones and other devices. Apple deployed 8,192 TPUv4 processors to power their server AI model.
Nvidia does not design TPUs, but instead focusses on so-called graphics processing units (GPUs), which are commonly employed in AI initiatives.
Unlike Nvidia, which sells its chips and systems separately, Google sells access to its TPUs via the Google Cloud Platform. Customers interested in purchasing access must create applications using Google’s cloud infrastructure in order to use the chips.
Apple is releasing pieces of Apple Intelligence to beta users this week.
Reuters reported on the use of the TPU chips in June, but Apple did not reveal the full degree of its reliance on Google hardware until Monday’s research study.
Google did not respond to a request for comment, while Nvidia declined to comment.
According to Apple’s developers, using Google’s chips would allow them to create even larger, more sophisticated models than the two detailed in the article.
Apple introduced a raft of new AI features at its developer conference in June, including the integration of OpenAI’s ChatGPT technology into its software.