Apple reportedly used Google chips to train two AI models

Apple has released the first version of Apple Intelligence, a collection of new AI functions for the company’s various operating systems that among other things will improve the voice assistant Siri and be able to generate text and images. (The new AI features, released Monday, are available in the developer betas for IOS 18.1 and iPad OS 18.1.)

Apple Intelligence is expected to be available to the public later this year, but is not expected to arrive at the same time as Apple’s new iPhone, expected to launch in September and initially run iOS 18.

At the same time,  Reuters reports, a new research report from Apple indicates the company will use Google’s, rather than Nvidia’s, hardware to train two of the central AI models. The decision is noteworthy because Nvidia currently makes the most sought-after AI processors.

Apple reportedly used 2048 of Google’s TPUv5 chips to build the AI ​​model to be used for the iPhone and 8192 TPUv4 processors for its server-based AI model. Apple itself has not yet commented on the matter.