Is Artificial Intelligence the New Meme Stock? Judging by the recent crazy share price of Arm Holdings plc, this is indeed the case.
The British semiconductor design company's share price doubled to $149 between Feb. 5 and Feb. 12. Since then, perhaps as the market has come to its senses, the stock price has fallen to $128. Arm's third-quarter earnings were the catalyst, with slightly better numbers but sparked widespread cheer about Arm's ability to harness the AI boom that brought chipmaker Nvidia CorpPushed to extremely high heights.
However, this should be taken with a grain of salt, because Arm is not directly involved in the hottest business at the moment: Nvidia's all-important artificial intelligence chips.
Nvidia's GPUs (Graphics Processing Units) were originally designed to render graphics in computer games, but over the years, they began to serve another purpose. Their superior parallel processing capabilities make them ideal for the computational needs of training and maintaining AI models. Today, GPUs are the standard for AI development, the equivalent of toilet paper during the pandemic. They're hard to find** and cost tens of thousands of dollars each. The insatiable demand for these chips has sent Nvidia's market capitalization soaring from $323 billion at the end of 2020 to a staggering $1$8 trillion. These chips can be approved by Amazon (Amazon.).com inc.) and Microsoft Corp) and other cloud providers to gain access.
Because of Arm's business of designing chips for data centers, it could still benefit from the AI boom, but it would be a pick-and-shovel success rather than the kind of immediate advantage Nvidia gets from AI-developed core sales products.
Cambridge, UK-based Arm does not make chips, but instead licenses its design blueprints, which are designed primarily for CPUs (** processors), which are more versatile processors that are used to perform many different tasks, such as running operating systems or managing data and network requests in data centers. Arm's design is not in the GPU because the chips are based on Nvidia's architecture. An ARM spokesman declined to comment.
Think of GPUs as high-paying sushi chefs that everyone wants to hire, and CPUs like restaurant managers. The latter doesn't directly address the food everyone is there, but it still benefits from the success of the GPU and everyone's desire for sushi (or AI).
For example, ARM can still benefit greatly from the broader demand for cloud companies that dominate the AI training business by powering complementary work in its data centers. In a partnership with Intel (Intel Corpand Advanced Micro Devices IncAfter years of competing with x86 processors, ARM's market share in the data center chip space has slowly grown to single digits. Arm's energy-efficient chip design makes it a staple of smartphones, which is partly a good score. At present, Arm's energy-efficient chip design has enabled large cloud computing companies such as Amazon to adopt Arm-designed GR**iton CPUs to power Amazon Web Services servers.
A spokesperson for Arm confirmed that the company does not break down the revenue it earns from different** sources such as data centers and smartphones, which makes it difficult to determine how much the company is really profiting from AI. Arm CEO Rene Haas said in a meeting with analysts earlier this month that "there's definitely growth on the data center side," but he's also a bit of a part of the AI hype.
For example, he mentions that 24,000 GH200 Grace Hopper chips from NVIDIA are using Arm's technology on a supercomputer in Germany, and then (perhaps opportunistically) that Arm sees "a growing demand for AI in data centers, both in terms of training and inference." ”
It's impressive that Arm's technology is in the middle of a German supercomputer, but it feels a bit far-fetched to tie it to today's broader need for AI model training and reasoning, even if it does mean Arm doing more restaurant managers. In most cases, the CPU will not perform this work for the AI model. Nvidia's GPUs can.
When Haas mentions that "more and more AI is running on more and more edge devices......And these devices are all running on ARM", which also seems a bit optimistic. "Edge devices" refer to smartphones that have their own AI chips that do not need to be connected to the internet to run services such as ChatGPT. But the market for such devices is still small and just getting started.
Another issue worth considering is that ARM is always faltering when it comes to entering new markets. In 2016, SoftBank Group CorpWhen founder Masayoshi Son bought Arm for $32 billion, he had big plans to dominate the burgeoning market for connected devices, known as the Internet of Things (IoT). But five years later, the company's IoT software sales have stagnated, a fraction of the company's goal of reaching $2 billion in revenue by 2025. Arm has been trying to pivot to a new business (software sales) that it is unfamiliar with, as for many years its main revenue has been collecting licensing and royalties for its mobile chip designs.
ARM could still benefit from the AI boom. The more people who want to eat sushi, the more people will patronize this restaurant. But despite the huge potential of the AI business, it may take a while for Arm to capitalize on its path, as it has done in the IoT space, and Arm will not have direct access to the core of the business – GPUs. This makes its current AI glow look less dazzling.