Avishek Das/SOPA Images/Sipa USA via AP
This week, the leaderboard for billionaire bragging rights got shaken up. Nvidia, the high-end semiconductor design company currently in demand for its specialization in chip processors for artificial intelligence, leapt past Microsoft, Amazon, and Apple in market capitalization to become the most valuable company in the world. (It dipped just below Microsoft on Thursday after falling 3.5 percent.) The company’s stock is worth around $3.33 trillion, after hitting the $2 trillion mark only four months ago, and $1 trillion only nine months before that.
The numbers are staggering. Nvidia started the year at the equivalent of $48.17 a share; it closed yesterday at $130.78. After the company announced first-quarter earnings in February, it had the largest one-day gain in Wall Street history. It is on track to become 15 percent of the S&P 500 all by itself, according to analysts at Evercore. The S&P has hit 31 record highs so far this year, with Nvidia the primary driver; the endless rise has actually pulled up other stock markets too. Once a relatively staid Silicon Valley company that didn’t offer free meals or a commuter benefit, Nvidia’s professional employees are now all millionaires. (Nvidia doesn’t make its own chips; that honor goes mostly to Taiwan Semiconductor Manufacturing Company, or TSMC.)
For investors, a bet on Nvidia is really a bet on AI. The company earned about $11 billion in revenue in 2020, mostly by selling graphics chips for video gaming. Its projected revenue in 2026 is nearly $131 billion, accomplished through discovering that its processors could power data center chips for large language models and other generative systems. Now it is the supplier to the AI revolution. Its earnings are still pretty low given the astronomical valuation, but as long as the market believes that AI is the only business development that matters in this century, then Nvidia will be the only stock that matters, at least until an OpenAI IPO.
There are obvious parallels to the dot-com bubble of the late 1990s, when any business on the internet was seen as a surefire winner. They weren’t, and neither is Nvidia. But much of this mania is built on Nvidia’s first-mover advantage.
There are other potential competitors for AI chips, like AMD and Intel. And customers of AI processors like Microsoft, Google, and Amazon have incentives to develop their own and stop paying an outside vendor. But Nvidia is far out in front; one estimate puts its market share of AI chips above 80 percent. Stock analysts openly enthuse about Nvidia’s lack of competition for AI computing. The company “continues to fundamentally widen [its] competitive moat,” said Vivek Arya of Bank of America in a research note in March.
The company is certainly primed to milk this monopoly. In March, CEO Jensen Huang said Nvidia’s latest AI chip would cost between $30,000 and $40,000 a unit. The premium on Nvidia’s graphic processors has led TSMC, the world leader for advanced semiconductors, to openly muse about raising its own prices. Nvidia accounts for 10 percent of TSMC’s revenue.
Huang very consciously developed an ecosystem for its products from which AI firms would be hard-pressed to extract themselves. Nvidia invested in over 30 startups just in 2023, including a cloud company called CoreWeave that uses Nvidia chips for its data centers. Nvidia also makes software used in creating AI applications that interact smoothly with the chips; there are indications that the company bundles the chip sales with the software.
The goal of all of this appears to be to impose a cost on bolting from Nvidia processors. That is an ever-present fear among the company’s customers, according to competitors. “A lot of people that we meet with say that if Nvidia were to hear that we were meeting, they would disavow it,” one chip startup CEO told The Wall Street Journal.
As big as Nvidia is now, it could have been bigger. The Federal Trade Commission successfully blocked Nvidia from acquiring Arm, a British chip designer that afterward did the largest IPO of 2023. FTC chair Lina Khan has taken credit for the fact that keeping the two companies separate raised the fortunes of both.
But now, Nvidia’s size is becoming a factor in its own right. Earlier this month, the FTC and the Justice Department reached a jurisdiction deal on antitrust investigations around AI firms. We are definitely seeing a rerun of the companies that dominated the Web 2.0 era having the data and corporate wealth to transfer into a domination of the AI future. CFPB director Rohit Chopra has called AI “a natural oligopoly,” and the antitrust authorities are poised to attack that.
This includes Nvidia itself. The Justice Department now has jurisdiction over the chipmaker, and the investigation has begun. (The FTC is looking at OpenAI and Microsoft.) A lot has been made of Big Tech’s practice of “strategic partnerships” in the AI space, which look like acquisitions in all but name. The Nvidia case looks like a more commonplace practice of locking in consumers amid threats of delays or denial of access if companies try to migrate off their systems.
Some industry observers believe that Nvidia’s days atop the AI compute list are numbered anyway. Short-seller interest in the company is nearly twice the level at Apple or Tesla. Many believe that, as computing power to train models becomes less important and less costly over time, and competitors catch up, Nvidia’s value could fade. Indeed, when companies made inroads into its former bread and butter of graphics chips, Nvidia dropped 50 percent or more in value 14 different times since it went public in 1999.
The bigger problem could be that AI is a bubble in and of itself, with far more funding committed than the technology’s value. McDonald’s giving up on using AI in drive-through ordering is an example of the limits to what the technology can accomplish. And for all the hype about large language models, startups and established companies alike have struggled to convert their attempts to use them into anything approaching profit, never mind anything resembling the right words and meanings. One estimate is that Nvidia has sold $50 billion in chips for AI, to companies that have only generated $3 billion in sales.
That raises questions as to whether the world’s largest company (or thereabouts) is built on a foundation of sand, and what happens should everyone realize that. But for now, the DOJ has a fairly simple job: learn more about whether Nvidia is bullying its customers into guaranteeing sales of its products, thereby stifling competition for AI chips and processing power.
Inevitably, the response from Nvidia and its allies will be that such a case would stunt the growth of AI and its impact on U.S. markets. But you could say that growing a single stock to the point where it is this critical to the functioning of the economy is precisely the problem.
