Apple did not use Nvidia GPUs to train its Apple Intelligence AI models.
NVIDIA (NVDA -0.21%) NVIDIA has been one of the best stocks to invest in since the beginning of 2023. The company’s stock price rally revolves around the massive demand for artificial intelligence (AI) computing power. Much of this demand comes from big technology companies, and one big technology company recently said no to NVIDIA.
Apple (AAPL 1.37%) Apple was a late entrant in the artificial intelligence space, but its products were highly anticipated. To the surprise of many, Apple did not use Nvidia’s GPUs (graphics processing units) to create its AI models. Is this a sign that Nvidia’s GPU empire is starting to crumble?
GPUs are the most popular choice for training AI models
Nvidia’s GPUs are extremely popular for running AI workloads. They’re best in class and have industry-leading software that helps you get the most performance out of your GPU. This is important because these servers that run AI models are equipped with thousands of GPUs. GPUs are ideal for these types of workloads because, unlike the central processing units (CPUs) found in PCs, they can perform many calculations in parallel. They’re also well-suited for other tasks like cryptocurrency mining, engineering simulations, and drug discovery.
But they weren’t purpose-built for AI calculations, and there are likely better ways to handle the calculations. These chips do exist, and they’re not made by Nvidia’s traditional competitors.
One such option is the alphabet (Google 0.95%) (Google 1.01%) Tensor Processing Units (TPUs). Alphabet’s TPUs are specifically designed for these large-scale AI calculations, but they are limited in their flexibility: With the right workload, TPUs can far outperform Nvidia’s GPUs, but only in specific use cases.
This is the main reason why Apple chose Alphabet’s TPUs to train its Apple Intelligence models. Does one of the industry’s biggest players choosing specialized hardware over broader hardware mean the end of Nvidia?
Nvidia’s investment thesis hinges on customers buying more GPUs
One of the reasons some investors (like me) are hesitant to buy Nvidia stock right now is that the company has inflated growth expectations built into it. Anyone buying the stock today is assuming that demand for Nvidia’s GPUs will continue to grow substantially in the future. But they’re not taking into account the possibility that once developers get used to configuring AI models in a certain way, they might choose to use specialized hardware like TPUs.
That could be a hypothetical challenge for Nvidia. Wall Street analysts expect the company’s revenue to grow 98% to $110 billion this year and 36% to $151 billion next year. Those projections could be in jeopardy if demand for the company’s products falls.
NVDA Revenue (TTM) data from YCharts. TTM = Trailing 12 Months.
Alphabet isn’t the only company to offer custom designs. Amazon (AMZN 0.69%) and Microsoft (MSFT 0.83%) Both companies have their own custom chips designed specifically for training AI models. None of these three companies sell these products directly to consumers. Instead, if clients want to access them, they must rent them through the companies’ cloud computing services.
This is a better business model in the long term than NVIDIA’s: When NVIDIA sells a GPU, the only revenue it gets comes from customers who decide to upgrade or replace the unit at the end of its useful life. Cloud computing companies buy hardware and then rent that power out to customers on a regular basis, expanding the value they get from a single unit.
Still, the big three cloud companies have data centers full of Nvidia GPUs and are buying more, so while there may be better ways to train AI models, many companies still train them on GPUs.
Only time will tell whether this trend reverses and more companies start training AI models like Apple has, or whether broader-purpose GPUs continue to dominate, but it’s an important point to keep in mind if you’re considering investing in Nvidia.
John Mackey, former CEO of Amazon subsidiary Whole Foods Market, is a member of The Motley Fool’s board of directors. Suzanne Frey, an executive at Alphabet, is a member of The Motley Fool’s board of directors. Keithen Drury has invested in Alphabet and Amazon. The Motley Fool has invested in and recommends Alphabet, Amazon, Apple, Microsoft, and NVIDIA. The Motley Fool recommends long January 2026 $395 calls on Microsoft and short January 2026 $405 calls on Microsoft. The Motley Fool has a disclosure policy.
