Why is the DOJ investigating Nvidia?

After a stock sell-off following its quarterly earnings report, Nvidia’s pain was aggravated last week by news that the Department of Justice is escalating its investigation into the company for anticompetitive practices.

According to a Bloomberg report, the DOJ sent a subpoena to Nvidia as part of a probe into alleged antitrust practices. Officials are said to be concerned that Nvidia makes it harder for customers to switch to other suppliers and penalizes buyers that do not exclusively use its AI chips.

However, analysts identify several problems with that argument. For one, Nvidia had no competition up until recently. AMD was fighting for its life and has only recently introduced a competitive AI product, the Instinct line. Intel is also very recent to market with Gaudi and GPU Max.

“I think it’s hard to make an anti-competitive case when they had no competition. They were the first to market. So how do you justify that when they’re the only ones out there with a solution?” said Jim McGregor, principal analyst with Tirias Research.

“Nvidia clearly has the power to misact, but opportunity alone doesn’t make for a crime, and the evidence of an actual crime appears to be mostly conjecture at this point. No smoking gun has yet emerged, and Nvidia states they haven’t even been notified of the investigation yet,” said Rob Enderle, principal analyst for The Enderle Group. 

Complaints are typically part of antitrust stirrings. In the 1980s, for example, Digital Research and IBM were very vocal about Microsoft practices shutting out their competitive products (DR-DOS and OS/2, respectively). And in the 1990s, Netscape was repeatedly complaining about Microsoft’s tactics in the browser market.

AMD – which has been gaining ground on Nividia with its Instinct line of GPU accelerators – has made no public complaints about Nvidia playing rough with customers. Nor has Intel or startups like SambaNova made such complaints, at least not publicly.

Nvidia does have its custom language, CUDA, which locks the customer into its hardware. But over time, that becomes less important because there are higher levels of abstraction when you’re dealing with libraries and models, said McGregor.

“When you’re dealing with Pytorch and stuff like that, CUDA doesn’t become as important, and you can easily cross compile from one platform to another, especially with the tools that are available now,” McGregor said.

Enderle said Nvidia’s status changed dramatically over the last year or so, as they took a massive lead in AI, which is likely why the regulators thought they needed to take a look. “The problem for Nvidia is that when groups like this look for problems, they generally find them. But there is no evidence of intent,” he said.

Another point in Nvidia’s favor is the fact that the entire industry is capacity constrained, since every GPU for the data center comes out of TSMC. Nvidia, AMD, and Intel have all said that they can’t get enough product out of TSMC, and that they would sell more if they could.

“It’s going to be hard to show anti-competitive practices when everybody in this market is capacity constrained, and everybody admits that they could ship more if they had more capacity, and all that capacity is from one producer, TSMC,” said McGregor.

Read more GPU news

• AMD doesn’t want to play games so much as it wants to embrace AI: A senior AMD executive spelled out what many have suspected: AMD is prioritizing AI performance over gaming performance as it joins other chip makers in chasing more AI business.
• Startups pursue GPU alternatives for AI: The GPU is the mainstay of AI processing, but several companies think they have a better option.
• Intel highlights new Xeons for AI at Hot Chips 2024: Intel’s photonics group demonstrated a new high-speed optical interconnect chiplet co-packaged with an Intel CPU and running live data at the show.
• 3 takeaways from AMD’s ZT Systems grab: AMD says the $4.9 billion deal to acquire the server maker is ‘our next major step’ in AI.
• Talent gap threatens US semiconductor industry: A shortage of skilled workers could affect the deployment of AI, which depends on chip innovation and semiconductor availability to thrive.