Google's custom chips were used to train Apple's AI models, according to the company.

• On Monday, Apple revealed in a technical paper that the two AI models powering its AI system, Apple Intelligence, were previously trained on Google-engineered chips in the cloud.
• Google's Tensor Processing Units, initially designed for internal tasks, are gaining wider acceptance.
• Nvidia's graphics processing units may no longer be the go-to choice for AI training as some big tech companies, including Apple, are exploring alternatives.

On Monday, it was reported that Apple Intelligence's AI models were pre-trained on processors developed by Google, indicating that big tech companies are exploring alternatives to NVIDIA when it comes to training advanced AI systems.

Apple published a technical paper detailing its use of a homegrown Tensor Processing Unit (TPU) for training, while also releasing a preview version of Apple Intelligence for some devices on Monday.

Nvidia's high-end AI training chips, which are known for their pricey graphics processing units (GPUs), have been in high demand over the past couple years, making it difficult to procure them in the quantities needed. OpenAI, Anthropic, Google, Microsoft, and other tech companies are all using Nvidia's GPUs for their models, while other companies are also using them to build out their AI systems and offerings.

Last week, both Mark Zuckerberg of Meta and Sundar Pichai of Alphabet acknowledged that their companies and others in the industry may be investing too much in AI infrastructure, but the risk of not doing so was too high.

Zuckerberg stated on a podcast with Emily Chang of Bloomberg that being behind in technology can put you out of position for the most important advancements for the next 10 to 15 years.

Apple's 47-page paper does not mention Google or Nvidia, but it states that its Apple Foundation Model (AFM) and AFM server were trained on "Cloud TPU clusters." This implies that Apple utilized servers from a cloud provider to execute the computations.

The system enables us to efficiently and scalably train AFM models, including AFM-on-device, AFM-server, and larger models, according to Apple's paper.

Representatives for Apple and Google didn't respond to requests for comment.

Apple was slower to unveil its AI plans compared to many of its competitors, who quickly adopted generative AI following the launch of ChatGPT in late 2022. On Monday, Apple unveiled Apple Intelligence, a system that includes new features such as a revamped Siri interface, improved natural language processing, and AI-generated summaries in text fields.

In the upcoming year, Apple intends to introduce features powered by generative AI, such as image and emoji generation, as well as an enhanced Siri that can utilize personal information and execute actions within applications.

According to Monday's paper, Apple revealed that AFM on-device was trained on a single "slice" of 2048 TPU v5p chips working together. This is the most advanced TPU, first launched in December. On the other hand, AFM-server was trained on 8192 TPU v4 chips that were configured to work together as eight slices through a data center network, as per the paper.

According to Google's website, the latest TPUs cost less than $2 per hour when booked for three years in advance. Google first introduced its TPUs in 2015 for internal workloads and made them available to the public in 2017. These chips are now considered one of the most mature custom chips designed for artificial intelligence.

Nvidia's top customer is still Google, which utilizes Nvidia's GPUs and TPUs for AI training and offers access to Nvidia's technology on its cloud.

Apple previously stated that inference, which involves using a pre-trained AI model to generate content or make predictions, would partially occur on Apple's own chips in its data centers.

Apple has published its second technical paper on its AI system, following a more general version released in June. The company stated at the time that it utilized TPUs while developing its AI models.

After the close of trading on Thursday, Apple will release its quarterly results.

WATCH: How the massive power draw of generative AI is overtaxing our grid