Sundar Pichai and Tim Cook
Source: Reuters; An apple
Apple said on Monday that the artificial intelligence models that underpin Apple Intelligence, its AI system, have been pre-trained on processors designed by Google, a sign that major tech companies are looking at alternatives to Nvidia when it comes to training cutting edge AI.
Apple’s choice of Google’s own Tensor Processing Unit (TPU) for training was detailed in a white paper just released by the company. Separately, Apple released a preview of Apple Intelligence for some devices on Monday.
Nvidia’s expensive graphics processing units (GPUs) dominate the market for high-end artificial intelligence training chips and have been in such demand for the past few years that they have been hard to come by in the required quantities. OpenAI, Microsoft and Anthropic use Nvidia GPUs for their models, while other tech companies including Google, Meta, Oracle and Tesla use them to build their AI systems and offerings.
Meta CEO Mark Zuckerberg and Alphabet CEO Sundar Pichai made comments last week suggesting their companies and others in the industry may be investing too much in AI infrastructure, but acknowledged the business risk of doing otherwise is too tall.
“The downside of being behind is that you’re out of position as the most important technology for the next 10 to 15 years,” Zuckerberg said in a podcast with Bloomberg’s Emily Chang.
Apple didn’t mention Google or Nvidia in its 47-page article, but noted that its Apple Foundation Model (AFM) and AFM server were trained on “Cloud TPU clusters.” This means that Apple has hired servers from a cloud provider to perform the calculations.
“This system allows us to train AFM models efficiently and scalable, including on-device AFM, AFM server, and larger models,” Apple said in the document.
Representatives for Apple and Google did not respond to requests for comment.
Apple was later to reveal its AI plans than many of its peers, which vociferously embraced generative AI soon after OpenAI’s ChatGPT launched in late 2022. On Monday, Apple unveiled Apple Intelligence. The system includes several new features, such as a refreshed Siri look, better natural language processing, and AI-generated summaries in text fields.
Next year, Apple plans to release features based on generative AI, including image generation, emoticon generation, and an enabled Siri that can access a user’s personal information and take actions inside apps.
In Monday’s article, Apple said the device’s AFM was trained on a single “slice” of 2,048 TPU v5p chips working together. This is the most advanced TPU, first released in December. The AFM server is trained on 8,192 TPU v4 chips that are configured to work together as eight segments across a data center network, according to the article.
Google’s latest TPUs cost less than $2 per hour of chip use when reserved three years in advance, according to Google’s website. Google first introduced its TPUs in 2015 for internal workloads and made them available to the public in 2017. They are now among the most mature custom chips designed for artificial intelligence.
Still, Google is one of Nvidia’s biggest customers. It uses Nvidia GPUs, its own TPUs to train AI systems and also sells access to Nvidia technology in its cloud.
Apple has previously said that inference, which means taking a pre-trained AI model and running it to generate content or make predictions, will partly happen on Apple’s own chips in its data centers.
This is the second technical document for Apple’s AI system, after a more general version was published in June. At the time, Apple said it was using TPU while developing its AI models.
Apple is scheduled to report quarterly results after the close of trading on Thursday.
