Apple has disclosed that it utilized Google’s custom chips to train the AI models that support Apple Intelligence. In a research paper released on Monday, the company revealed that two crucial components of its AI system were pretrained using processors designed by Google. This approach is unusual given the heavy reliance of major tech firms on AI chipmaker Nvidia.

The company revealed that it utilized two versions of Google's custom tensor processing units (TPUs) organized into large clusters. For training the AI model used in iPhones and other devices, Apple employed 2,048 TPUv5p chips. Meanwhile, for its server-based AI model, Apple utilized 8,192 TPUv4 processors.

Additionally, the iPhone maker has released a preview of its much-anticipated Apple Intelligence system.

Nvidia currently dominates approximately 80% of the AI chip market with its high-demand, costly graphics processing units (GPUs), which have surged in popularity since the AI boom.

Recently, several tech giants, including Google, Meta, Microsoft, and Amazon, have started developing their own AI chips due to the high costs associated with training large-scale AI models.

Google’s TPUs are available through Google Cloud Platform, which means users need to develop their software on Google Cloud to utilize these chips.

The research paper also mentioned that Google's chips could enable the creation of even larger and more complex AI models beyond those currently outlined.

At its WWDC conference in June, Apple introduced a range of new AI tools and features, including the integration of OpenAI’s ChatGPT into its software. Selected Apple Intelligence features are expected to be rolled out to beta users later this week.