So Google's choice of processors, specifically tensor processing units (TPUs), over Nvidia's GPUs has been a strategic move to develop advanced AI capabilities for devices like iPhones, iPads, and Macs.
Technical Questions
The leaked documents indicate that Apple chose to use 2,048 TPUv5p chips from female database Google for AI models used in personal devices. For models operated on servers, 8,192 TPUv4 processors were used.
As such, Apple's decision to use Google's TPUs highlights how affordable and technically capable these chips are for complex AI tasks.
Additionally, Google's TPUs, initially developed for internal use and later made available to the public, are known for their efficiency and cost-effectiveness.
Google offers these processors through its Cloud Platform, allowing developers to build and train AI models directly on Google's cloud infrastructure.
This approach facilitates access to advanced technologies without the need for large investments in proprietary infrastructure.
Reasons behind Apple's choice
Apple’s decision not to use Nvidia GPUs, which dominate the AI hardware market with about 80% market share, is significant. While Nvidia GPUs are widely recognized for their performance in AI tasks, Apple decided to explore other options that could offer specific advantages in terms of cost and integration capabilities with its systems.
In addition to efficiency, the choice of Google TPUs may be related to cost issues. This is because using Nvidia GPUs can be financially more expensive, especially on a large scale. By opting for Google TPUs, Apple is able to maintain a balance between performance and cost, optimizing its investments in AI technology.
Another important consideration is the flexibility of Google’s TPUs. The developers designed these processors to be highly scalable, allowing Apple to adapt their processing capabilities as needed. This is particularly important in a field as dynamic as artificial intelligence, where processing demands can vary significantly.