Nvidia has a China problem. Chinese companies are some of Nvidia’s biggest clients, with the country traditionally contributing about 20% of the company’s revenue. Yet the Biden administration is worried that the company’s chips, key to the AI revolution, are a national security risk if sent to China. That’s giving the world’s most valuable chip company a dilemma: how to stay involved in the world’s second-largest economy while still complying with Washington’s rules.
On Monday, Nvidia announced that four car manufacturers—Li Auto, Great Wall Motor, ZEEKR, and Xiaomi—will use the company’s DRIVE technology to power their automated driving systems.
While known for its graphics processing units (GPUs), which are now key to AI-related applications, Nvidia is also trying to get a foothold in the auto industry. The company has announced car-focused partnerships with both chip designer MediaTek and electronics manufacturer Foxconn. Nvidia CEO Jensen Huang even joined Foxconn’s “Tech Day” to announce the chipmaker would be a key partner in the Apple supplier’s efforts to start making electric cars.
Complying with U.S. rules
Nvidia is also planning to start mass production of a new set of semiconductors designed for the China market that comply with updated U.S. export restrictions, reports Reuters. The company reportedly plans to start production later this year.
“We are engaged with the U.S. government and, following its clear guidelines, are working to offer compliant product solutions to customers worldwide,” an Nvidia spokesperson said to Fortune in a statement.
This isn’t the first time Nvidia has had to change its products to comply with U.S. rules. After the U.S. announced its first controls on semiconductor sales to China in October 2022, Nvidia developed chips for the Chinese market that offered similar capabilities to its high-end chips while still complying with U.S. rules.
Last October, the Biden administration updated its export restrictions to close this loophole, forcing Nvidia to redevelop its products again. The updated rules even prevented Nvidia from selling its most advanced video gaming chips to Chinese firms. The chipmaker released a U.S.-compliant replacement in December, which it said was developed after the company “extensively engaged with the U.S. government.”
Nvidia CEO Jensen Huang has, at times, warned against U.S. rules against chip sales to China. He previously warned, in an interview with the Financial Times in May, that U.S. export controls aimed at slowing China’s advance in the semiconductor industry would instead encourage the country to build its own chips in a bid to rival Nvidia’s. Then in another interview with CNBC in November, Huang said U.S. chipmakers were at least a decade away from “supply-chain independence” from China.
Nvidia chief financial officer Colette Kress has downplayed the near-term effect of U.S. export restrictions, yet warned that, in the long term, U.S. restrictions would lead to a “permanent loss of opportunities for the U.S. industry to compete and lead in one of the world’s largest markets.”
Last quarter, the company warned that data center revenue from countries affected by U.S. controls, including China, will “decline significantly.”
Customer demand
Yet Nvidia has a new problem: Chinese companies which originally flocked to its high-powered chips aren’t as interested in weaker replacements.
Local Chinese chipmakers are pitching their own chips as a safer alternative, banking on fears among China’s tech companies that the U.S. will once again tighten restrictions, according to Reuters in December.
Companies are also concerned that Nvidia’s downgraded processors may mean there’s not as much difference in performance compared with those produced by Chinese manufacturers, the Wall Street Journal reports. Firms like Alibaba and Tencent are now shifting some orders to Chinese companies like Huawei; the Chinese tech company has a new AI chip that Reuters describes as comparable to Nvidia’s offerings.
Update Jan. 9, 2023: This article has been updated with a comment from Nvidia.
