Unlike most companies that develop their own artificial intelligence models, Apple has decided to use Google chips rather than NVIDIA. This was reported by Reuters.

This information emerged from Apple’s reports published on July 29, 2024. According to them, the company used two types of Google’s Tensor Processing Unit (TPU) to train its models, which are organized into large clusters of chips.

The company even specified how many chips were used and for what purposes. The company used 2048 TPUv5p chips to train the AI model that will run directly on iPhones and other devices. For the Apple Intelligence server model, 8192 TPUv4 processors were used.

The decision not to use the power of NVIDIA GPUs is quite interesting. The company’s report does not specify that the company did not use such processors in any way, but there is no mention of NVIDIA either.

Among other things, unlike NVIDIA, Google sells access to its TPUs through the Google Cloud Platform. Customers who want to use this solution must create their software through the Google Cloud Platform.