Apple is using Google's TPUs instead of Nvidia GPUs to train its artificial intelligence (AI) models.
In a recent research paper published by Apple, the company's hardware and software infrastructure descriptions for its AI tools omit Nvidia chips.
Throughout the paper, Apple details that it instead used Google's Tensor Processing Units (TPUs) to train its AI models.
The company used 2,048 TPUv5p chips for its AI models for iPhones and other devices, and 8,192 TPUv4 processors for its server AI model.
TPUs are accessible via the Google Cloud Platform and are Google's custom chips. In May 2024, Google revealed its sixth generation of its TPUs, dubbed Trillium, claiming that it has a 4.7x compute boost and is 67 percent more energy efficient than the previous generation.
Apple’s engineers noted that Google’s chips could potentially facilitate the creation of even larger, more sophisticated models than those discussed in their paper.
Apple's AI offering - called Apple Intelligence - will be available for testing by its beta users this week.
Apple has long been a Google Cloud customer and is believed to be the platform's single biggest storage customer. The company is reportedly working with TSMC to develop its own chips.
Reuters first reported that Apple was using Google to help train its AI models in June 2024, though the seeming absence of Nvidia chips was only revealed in Apple's latest paper.
Nvidia controls around 80 percent of the market for AI chips. The company has seen massive growth over the last year, having reached a $3.34 trillion valuation in June 2024.
This followed the company hitting $2 trillion in February 2024, after a $1 trillion valuation in May 2023.
Despite this, shares in Nvidia have taken a hit this month, with the company's value dropping by 6.8 percent on July 24 alone. Some of this has been put down to concern about the revenue generated by AI.
Earlier this month, a research newsletter from Goldman Sachs warned that, despite large tech companies investing massive amounts in fueling generative AI, they have yet to demonstrate sustainable business models.
The investment banking firm estimates that around $1 trillion will be spent on AI infrastructure in the next year, but says that even if a 'killer application' emerges, it is unclear if it will generate the financial returns.
Google similarly saw a hit to its share value, having dipped by 7.24 percent this month.