Apple Leans on Google’s AI Chips for Its Long-Awaited AI System, "Apple Intelligence"
Apple’s long-awaited entry into the artificial intelligence (AI) race, Apple Intelligence, has sparked a wave of interest in the tech world, and new details reveal a surprising partnership with Google. Apple’s technical paper, published alongside the preview release of Apple Intelligence, has disclosed that the company used Google’s Tensor Processing Units (TPUs) to train its Apple Foundation Model (AFM). This move signifies a major deviation from the widespread reliance on Nvidia’s GPUs for AI training, raising questions about the future of the AI chip market.
Key Takeaways:
- Apple relied on Google’s TPUs, not Nvidia’s GPUs, to train its foundation model for Apple Intelligence. This shift reveals a divergence from industry trends and highlights Google’s growing prominence in the AI chip space.
- This decision could shake up the AI chip market, where Nvidia currently dominates. The move underscores the importance of having alternative AI training solutions.
- Google’s TPUs are now considered among the most mature custom chips designed for AI, highlighting its commitment to the field alongside its core search engine business.
- Apple’s use of Google’s TPUs while simultaneously promoting its own silicon advancements indicates the growing complexity and competition in the AI infrastructure landscape.
A Shift in the AI Chip Landscape
The AI industry’s dominance by Nvidia has been challenged, as seen in Apple’s decision. Nvidia’s GPUs have become the go-to choice for training large language models like OpenAI’s ChatGPT, driving a surge in demand that has led to shortages and exorbitant prices. But Apple’s reliance on Google’s TPUs for Apple Intelligence highlights the emergence of robust alternatives.
Google’s TPUs, first introduced in 2015, have evolved significantly and are now a strong contender in the AI chip market. Google’s commitment to developing custom AI hardware parallels its aggressive advancement in AI research and products. The company has been actively pushing its cloud computing services, including access to its TPU infrastructure, demonstrating its aspiration to become a dominant player in the AI space beyond its core search business.
Apple’s Strategic Move
Apple’s choice of Google’s TPUs for AI training is a strategic one. While Apple has been investing heavily in its proprietary silicon for its devices, the company recognizes the need to leverage the best resources for AI development, especially during the initial stages of training its foundation model. Google’s TPUs offer both scalability and efficiency, enabling Apple to train its models on a massive scale.
Furthermore, Apple’s reliance on Google’s TPUs doesn’t negate its own efforts in silicon development. Apple’s announcement that it will use its own chips for "inferencing" — the process of using trained models — aligns with its goal of retaining control and optimizing AI experiences for its users.
Implications for the Future of AI
The tech world is witnessing a rapid evolution in the AI landscape, and the relationship between Apple and Google underscores this dynamic. Apple’s adoption of Google’s TPUs, a move towards collaboration instead of direct competition, suggests a potential shift in how companies approach AI development. This shift could lead to a more diverse and robust ecosystem of AI infrastructure, offering greater choice and potentially pushing innovation.
The growing competition in the AI chip market could further fuel innovation and lower costs in the long run. As companies like Google and Apple invest in their own hardware solutions and explore partnerships with other players, the future of AI infrastructure remains an exciting and evolving landscape to watch.
