Instead of using Nvidia chips, Apple uses Google Tensor chips to develop Apple Intelligence.
Apple Intelligence is developed using Google hardware. Photo: Shutterstock.
Apple on July 30 suddenly announced that the Apple Intelligence tool has been pre-trained on Google-designed processors.
This information was shared by Apple in a technical document posted on its website. Specifically, instead of GPUs from Nvidia, Apple has used Google Tensor processor (TPU) to train and deploy the Apple Intelligence model.
Also in the 47-page report, the iPhone manufacturer revealed that cloud processing (AFM) features and AFM servers are trained on the TPU cloud cluster. This means that Apple has rented servers from the cloud service provider to perform the calculations.
“This system allows us to train AFM models efficiently and scalably, including on-device AFM, server-based AFM, and larger models,” Apple wrote in the report.
CNBC said this is a sign that tech giants are looking for alternatives to Nvidia hardware in training advanced AI models.
Nvidia is considered the most successful name thanks to the explosion trend of artificial intelligence (AI). According to Mizuho Securities, the company controls 70-95% of the AI chip market, which is used to train and deploy models such as OpenAI GPT.
As the training phase passes, companies turn to investing in systems that implement AI, also known as inference, which requires less power and cost. For comparison, Nvidia high-end AI chips cost around $30,000, making many customers eager to find an alternative.