(Bloomberg) — The U.S. Justice Department has sent subpoenas to Nvidia Corp. (NVDA) and other companies as it seeks evidence that the chipmaker violated antitrust laws, an escalation of its investigation into the dominant supplier of artificial intelligence processors.
Most read on Bloomberg
The Justice Department, which had previously sent questionnaires to companies, is now sending legally binding requests that compel recipients to provide information, according to people familiar with the investigation. This brings the government one step closer to filing a formal complaint.
Antitrust officials are concerned that Nvidia is making it harder to switch to other suppliers and penalizing buyers who don’t exclusively use its artificial intelligence chips, according to the people, who asked not to be identified because the discussions are private.
Nvidia shares, which suffered a record drop on Tuesday, fell further in late trading after Bloomberg reported the subpoenas. Still, the stock has more than doubled in value this year, boosted by explosive sales growth at the Santa Clara, California-based chipmaker.
As part of the investigation, which Bloomberg previously reported in June, investigators have been reaching out to other tech companies for information. The Justice Department’s San Francisco office is leading the probe, the people said. A Justice Department representative declined to comment.
In response to questions about the research, Nvidia said its market dominance comes from the quality of its products, which offer faster performance.
“Nvidia wins on merit, as reflected in our benchmark results and the value to customers, who can choose the solution that best fits their needs,” the company said in an emailed statement.
Nvidia has been under regulatory scrutiny since it became the world’s most valuable chipmaker and a key beneficiary of booming artificial intelligence spending. Sales have more than doubled each quarter and it has eclipsed onetime chip leaders such as Intel Corp.
As part of the Justice Department investigation, regulators have been looking into Nvidia’s acquisition of RunAI, a deal announced in April. That company makes software to manage artificial intelligence computing, and there are concerns that the merger will make it harder for customers to switch from Nvidia chips. Regulators are also looking into whether Nvidia offers preferential supply and pricing to customers who use its technology exclusively or buy its complete systems, according to the people.
Nvidia, founded in 1993, made its name selling graphics cards to gamers. But its chip-making method proved useful for creating AI models, a process that involves bombarding software with data. The company has also rapidly expanded its offerings with a range of software, servers, networking and services, all aimed, Nvidia says, at accelerating the deployment of AI.
Nvidia CEO Jensen Huang said it prioritizes customers who can use its products in ready-to-use data centers as soon as it provides them, a policy designed to avoid warehousing and accelerate broader adoption of AI.
The success of its products, coupled with rivals’ struggles to offer alternative chips, has made Nvidia a crucial part of the supply chain for some of the world’s largest companies. Microsoft Corp. and Meta Platforms Inc., for example, spend more than 40% of their hardware budgets on equipment from the chipmaker. At the peak of shortages for Nvidia’s H100 accelerator, individual components were retailing for as much as $90,000 each.
Analysts expect Nvidia to make $120.8 billion in revenue in 2024, up from $16 billion in 2020, with most of that money coming from its data center unit. In fact, Nvidia is expected to make more profit this year than the total sales of its closest rival, Advanced Micro Devices Inc.
There are also broader regulatory questions about Nvidia’s practices. Access to AI capabilities has become a key issue for governments around the world as the technology becomes increasingly vital to economic strength and national security.
(Updated with Nvidia’s response from the sixth paragraph.)
Most read from Bloomberg Businessweek
©2024 Bloomberg LP