$Apple (AAPL.US)$ has disclosed that the artificial intelligence models powering its AI system, Apple Intelligence, were trained on custom chips developed by $Alphabet-C (GOOG.US)$'s Google, not on the widely used chips from $NVIDIA (NVDA.US)$.
What Happened: Apple's AI models were trained on Google's Tensor Processing Unit (TPU), as per a technical paper released by the company on Monday. This announcement suggests that major tech companies are exploring alternatives to Nvidia for the training of cutting-edge AI.
Nvidia's GPUs have been the dominant choice for high-end AI training chips, but their high demand has made them challenging to obtain in the required quantities. Despite this, tech giants like OpenAI, $Microsoft (MSFT.US)$, and Anthropic continue to use Nvidia's GPUs for their models.
Other companies, including Google, $Meta Platforms (META.US)$, $Oracle (ORCL.US)$, and $Tesla (TSLA.US)$, are also acquiring these GPUs to expand their AI systems and offerings.
Last week, Meta CEO Mark Zuckerberg and Alphabet CEO Sundar Pichai both indicated that their companies, along with others in the industry, might be overinvesting in AI infrastructure. However, they emphasized that the business risk of underinvesting was too significant to ignore.
Apple's decision to use Google's TPUs for training is a significant move, given the industry's reliance on Nvidia's GPUs. This choice could potentially impact the AI chip market and the competition among major tech companies.
Apple's 47-page paper doesn't specifically mention Google or Nvidia, but it does highlight that its Apple Foundation Model (AFM) and AFM server are trained on "Cloud TPU clusters." This indicates that Apple utilized rented servers from a cloud provider for the necessary computations, according to a CNBC report.
Apple did not immediately respond to Benzinga's request for comment.
Why It Matters: This development comes amid a series of significant events in Apple's AI journey. The company recently announced a delay in the launch of its AI features, called Apple Intelligence, from September to October. These features were set to debut with iOS 18 but were postponed to ensure stability and performance for the new generative AI capabilities.
Apple's AI features are expected to be available on the latest devices, including the iPhone 15 Pro and Mac computers with M1 chips. The company also plans to integrate its A18 chip into its upcoming entry-level iPhones and iPads to enable them to run Apple Intelligence.
Ahead of its third-quarter earnings report on Aug. 1, Bank of America analyst Wamsi Mohan reiterated a Buy rating for Apple, maintaining a price target of $256. In a note published Monday, Mohan forecasted third-quarter revenue of $84.5 billion and EPS of $1.35, slightly above Wall Street estimates of $84.4 billion and $1.34, respectively.
The analyst anticipates that investors will closely scrutinize management's commentary on AI, Chinese market share, and services revenue.
This story was generated using Benzinga Neuro and edited by Kaustubh Bagalkote