Apple has diverged from industry norms by using Google’s tensor processing units (TPUs) rather than Nvidia’s GPUs for training two of its major AI models, according to a research paper published on Monday. This move highlights a significant shift in the tech giant’s approach to building its AI infrastructure.
The research paper reveals that Apple utilized Google’s TPUs, organized in extensive clusters, to develop key components of its upcoming AI tools and features. This decision is notable given Nvidia’s dominance in the AI hardware market, where it holds approximately 80% of the share with its highly sought-after processors.
Apple’s choice to rely on Google’s cloud-based hardware is intriguing, as it diverges from the prevalent industry trend where Nvidia’s chips are typically favored for their advanced capabilities in AI and machine learning tasks. The research paper, however, does not explicitly state that Nvidia chips were entirely excluded from Apple’s AI training process. Instead, the paper’s detailed description of Apple’s hardware and software setup does not include any mention of Nvidia’s products.
Nvidia’s GPUs are widely recognized for their powerful performance in AI computations, making Apple’s choice of Google’s TPUs a notable departure from the norm. This move could reflect strategic decisions related to cost, performance, or potential collaborations with Google.
In addition to its AI developments, Apple continues to expand its technological capabilities and infrastructure, aiming to enhance its suite of AI tools and features in future releases. As the AI landscape evolves, Apple’s decision underscores a growing trend of companies exploring diverse technological partnerships and solutions.
