What you need to know about Nvidia and the AI chip arms race – Marketplace
While Nvidia’s share price is down from its peak earlier in the week, its stock has skyrocketed by 262% in the past year, going from almost $242 a share at closing to $875.
The flourishing artificial intelligence industry has accelerated demand for the hardware that underpins AI applications: graphics processing units, a type of computer chip.
Nvidia is the GPU market leader, making GPUs that are used by apps like the AI chatbot ChatGPT and major tech companies like Facebook’s parent company, Meta.
Nvidia is part of a group of companies known as The Magnificent Seven, a reference to the 1960 Western film, that drove 2023’s stock market gains. The others in that cohort include Alphabet, Amazon, Apple, Meta, Microsoft and Tesla.
But Nvidia faces competitors eager to take a share of the chip market and businesses that want to lessen their reliance on the company. Intel plans to launch a new AI chip this year, Meta wants to use its own custom chip at its data centers and Google has developed Cloud Tensor Processing Units, which can be used to train AI models.
There are also AI chip startups popping up, which include names like Cerebras, Groq and Tenstorren, said Matt Bryson, senior vice president of research at Wedbush Securities.
GPUs were originally used in video games to render computer graphics, explained Sachin Sapatnekar, a professor of electrical and computer engineering at the University of Minnesota.
“Eventually, it was found that the kinds of computations that are required for graphics are actually very compatible with what’s needed for AI,” Sapatnekar said.
Sapatnekar said AI chips can do parallel processing, which means they process a large amount of data and handle a large amount of computations at the same time.
In practice, what that means is AI algorithms now have the capability to train on a large number of pictures to figure out how to, say, detect whether an image of a cat is of a cat, Sapatnekar explained. When it comes to language, GPUs help AI algorithms train on a large amount of text.
These algorithms can then in turn produce images resembling a cat or language mimicking a human, among other functions.
Right now, Nvidia is the leading manufacturer of chips for generative AI and it’s a very profitable company, explained David Kass, a clinical professor at the University of Maryland’s Robert H. Smith School of Business.
Nvidia has 80% control over the entire global GPU semiconductor chip market. In its latest earnings report, Nvidia reported a revenue of $22.1 billion for the fourth quarter of fiscal year 2024, which is up 265% since last year. Its GAAP earnings (earnings based on uniform accounting standards and reporting) per diluted share stood at $4.93, which is up 765% since last year. Its non-GAAP earnings (which exclude irregular circumstances) per diluted share was $5.16, an increase of 486% compared to last year.
Another reason Nvidia’s share price may have skyrocketed in recent months is because the success of the stock itself is attracting additional investment, Kass said.
Kass explained individuals and institutions may be jumping on the train because they see it leaving the station. Or, in other words: FOMO, he said.
Bryson of Wedbush Securities pointed out that the company was also able to differentiate itself through the development of CUDA, which Nvidia describes as a “parallel computing platform and programming model.”
Nvidia’s success doesn’t necessarily mean that its GPUs are superior to the competition, Bryson added. But he said the company has built a powerful infrastructure around CUDA.
Nvidia has developed its own CUDA programming language and offers a CUDA tookit that includes libraries of code for developers.
“Let’s say you want to perform a particular operation. You could write the code for the entire operation from scratch. Or you could have specialized code that already is made efficient on the hardware. So Nvidia has these libraries of kind of pre-bundled packages of code,” Sapatnekar said.
With Nvidia far ahead of the competition, Bryson said Advanced Micro Devices, or AMD, is trying to stake a position as the second-leading player in the AI chip space. AMD makes both central processing units, competing with the likes of Intel, and GPUs.
AMD share price has risen by about 143% since last year as demand for AI chips has grown.
Jeffrey Macher, a professor of strategy, economics and policy at Georgetown University’s McDonough School of Business, said he questions whether Nvidia will be able to meet all of the rising demand for AI chips on its own.
“It’s going to be an industry that’s going to see an increased number of competitors,” Macher said.
Despite the success of Nvidia and AMD, there are wrinkles in their supply chains. Both rely heavily on Taiwan Semiconductor Manufacturing Co. to make their chips, which will leave them vulnerable if anything goes awry with the company.
Macher said the semiconductor market used to be vertically integrated, meaning the chip designers themselves manufactured these chips. But Nvidia and AMD are fabless companies, which means they’re companies that outsource their chip manufacturing.
As we saw during the early stages of the COVID-19 pandemic, supply chain disruptions led to shortages across all kinds of different sectors, Marketplace’s Meghan McCarty Carino reported.
TSMC is planning to build Arizona chip plants which may help alleviate some of these concerns. But tech publication The Information reported that these chips “will still require assembly in Taiwan.”
And TSMC’s location carries geopolitical risks. If China invades Taiwan and TSMC becomes a Chinese company, U.S. companies may be reluctant to use TSMC out of fear that the Chinese government will appropriate their designs, Macher said.
Kass said he doesn’t see similarities between Nvidia’s rising stock and the dot-com bubble in the early 2000s, when many online startups tanked after their share prices reached unrealistic levels thanks to an influx of cash from venture capital firms that were overly optimistic about their potential.
Kass said some of these companies not only failed to make a profit, but weren’t even able to pull in any revenue either, unlike Nvidia, which is backed by real earnings.
He does think there could be a correction or a point where Nvidia stock will be perceived as overvalued. He explained the larger your company, the more difficult it is to sustain your rate of growth. Once that growth rate comes down, there could be a sharp sell-off.
But Kass said he doesn’t think there will be a sustained and/or a steep downturn for the company.
However, AI’s commercial viability is uncertain. Bryson said there are forecasts of how large the AI chip market will become – AMD, for example, suggested that the AI chip market will be worth $400 billion by 2027 — but it’s hard to validate those numbers.
Bryson compared AI with 4G, the fourth generation of wireless communication. He pointed out that apps like Uber and Instagram were enabled by 4G, and explained that AI is similar in the sense that it’s a platform that a future set of applications will be built on.
He said we’re not really sure what many of those apps will look like. When they launch, that will help people better assess what the market should be valued — whether that’s $400 billion or $100 billion.
“But I also think that at the end of the day, the reason that companies are spending so much on AI is because it will be the next Android or the next iOS or the next Windows,” Bryson said.
There’s a lot happening in the world. Through it all, Marketplace is here for you.
You rely on Marketplace to break down the world’s events and tell you how it affects you in a fact-based, approachable way. We rely on your financial support to keep making that possible.
Your donation today powers the independent journalism that you rely on. For just $5/month, you can help sustain Marketplace so we can keep reporting on the things that matter to you.
