Apple recently unveiled its AI system, Apple Intelligence, featuring generative AI models trained for various tasks like writing, summarizing, and image creation. Interestingly, these models weren’t trained using the popular Nvidia GPUs, but rather Google’s Tensor Processing Units (TPUs). This decision by Apple, one of the world’s wealthiest companies, raises questions about Nvidia’s stronghold in the AI chip market.
The Apple Foundation Model: Built on Google TPUs
Apple chose Google’s fourth and fifth-generation TPUs to train its foundation models for Apple Intelligence. TPUs, as application-specific integrated circuits (ASICs), are designed to efficiently run specific processes, making them a cost-effective alternative to GPUs for training AI models.
Google claims its TPU v5p is significantly faster than its previous generation, and even on par with Nvidia’s latest H100 chips. With Google’s sixth-generation TPU promising even more speed improvements, Apple’s decision to use TPUs seems like a strategic move to achieve cost efficiency without compromising performance.
A Warning Sign for Nvidia Investors?
While Apple isn’t the only company utilizing Google TPUs, its choice highlights the growing competition faced by Nvidia in the AI chip market. Major tech companies like Microsoft, Amazon, and Meta, who are also Nvidia’s biggest customers, are developing their own custom silicon for AI training and running.
Alphabet CEO Sundar Pichai acknowledged the current trend of heavy investment in AI chips and data centers, emphasizing the potential losses from under-investing. However, Google TPUs and other custom silicon might prove more cost-effective in the long run, as Apple’s decision demonstrates.
Customer Concentration and Nvidia’s Future
Nvidia heavily relies on a few tech giants for a significant portion of its revenue. With these companies striving to reduce their dependence on Nvidia, the chip designer could face slower revenue growth in the coming years.
The high expectations surrounding Nvidia, reflected in its current share price, might be challenged if more AI developers opt for cost-effective ASICs over Nvidia GPUs. A single disappointing quarter could significantly impact the stock price.
Nvidia’s AI Dominance: Not Forever
Nvidia’s current valuation, trading at a high price-to-earnings ratio, indicates the market’s optimism about its continued dominance in the AI chip market. However, the potential for a slowdown in sales and profits due to increased competition and the growing popularity of alternative solutions like Google TPUs poses a significant risk for Nvidia’s future.
Add Comment