Apple Uses Google Chips to Train AI: What Does It Mean for Nvidia?
Date: 2024-08-12 Author: Gabriel Deangelo Categories: BUSINESS
When Apple decided to train its AI models, it had the same options as other deep-pocketed tech giants. It could use its own chips, train them on cloud infrastructure, or, like many others, spend billions of dollars to buy Nvidia chips. But Apple chose a different path and struck a deal with the company that owns Google.
Some say Apple is late to the AI party. While its competitors have been pushing generative AI and investing billions of dollars in training and cloud services, Apple has focused on iPhone sales and “spatial computing.” That’s raised concerns among analysts: Is the company founded by Steve Jobs falling behind its closest competitors?
In June 2024, the fears seemed justified when Nvidia overtook Google, Apple, and Microsoft to become the most valuable company in the world.
Nvidia's Dominance
At the peak of its market value, around June 19, 2024, Nvidia's market cap reached an impressive $3.45 trillion, making it not only the market leader in 2024, but also the most valuable company in history. Even more astonishing was the fact that its value had grown by 147% in a year and by 2,617% over the past five years.
Economists and analysts have pointed to many factors driving the company's growth, but Nvidia's main source of revenue since 2017 has remained graphics processing units (GPUs). By the end of 2023, the company's products were estimated to account for about 70% of all AI chip sales.
Apple's Closed Ecosystem
Apple has always been keen to keep its plans secret. Although the company is public and required to disclose certain information, it has always had a unique approach to advertising, marketing, and setting consumer expectations.
However, a newly published study has revealed the secrets behind the development and training of Apple’s “Fundamental Language Models” that power “Apple Intelligence.” Instead of using Nvidia’s popular H-100 chips or developing its own new chip, the Cupertino company turned to Google.
According to the study, the new model was trained on more than 10,000 Google Tensor Processing Units (TPUs), most of which were TPUv4, with about 2,000 TPUv5.
While one might speculate that the choice was made due to the difficulty and high cost of acquiring Nvidia’s chips, there’s no doubt that Apple, with a market cap of $3.28 trillion as of August 10, 2024, could afford to acquire the market leader’s chips.
Neglect or Strategy?
It’s hard to pinpoint exactly why Apple has passed over Nvidia. Research shows that GPUs can offer more power and efficiency in certain areas, but Apple’s specialized approach seems to better leverage Google’s TPUs.
The big question: What does this mean for Nvidia? After skyrocketing its market cap from $2 trillion to $3 trillion, the company’s stock has begun to correct to a more stable level.
Apple’s choice could signal a change to come, as more major tech companies begin to develop their own chips. The overall rise of interest in generative AI could work against Nvidia, with specialized chips replacing less tailored solutions.
Or is this just another example of Apple forging a path different from its competitors. Companies like OpenAI, Microsoft, and Nvidia don’t have the same broad user base as Apple.
Prioritizing billions of iPhone users over corporate trends and risky bets on artificial general intelligence has kept Apple’s shareholders happy for decades. It's unlikely that this will change until consumer spending on AI services catches up with the industry hype.
