Nvidia Corp faces new US restrictions on the export of its H20 chip to China, a policy change that will cost the company billions of dollars and hamstring a product line that it explicitly designed to comply with previous curbs.
The government informed Nvidia on Monday that the H20 would require a license to export to China “for the indefinite future,” the company said in a regulatory filing Tuesday.
Officials said that the new rules address concerns that “the covered products may be used in, or diverted to, a supercomputer in China,” according to the filing.
Nvidia warned that it will report about US$5.5 billion in writedowns during the fiscal first quarter from “inventory, purchase commitments and related reserves” tied to the H20 line.
The company’s shares slid about 6% in late trading following the announcement. Advanced Micro Devices Inc, which competes with Nvidia in the AI chip market, slumped as well.
A representative for the White House didn’t immediately respond to a request for comment.
See also: TSMC, ASML outlooks to reveal depth of tariff pain, AI angst
Bloomberg News reported in January that the Trump administration was exploring such a step. Though the H20 can be used to develop and run artificial intelligence software and services, it’s a scaled-down product specifically designed not to be too powerful.
The H20 had allowed the company to continue serving the market for data centre AI chips in that country, though the product isn’t as fast as Nvidia’s non-China offerings for the training of models. It was better-suited to the inference stage — the point where the AI model recognizes patterns and draws conclusions.
Now it’s also seen as potentially too risky to export to China, the main US rival in artificial intelligence and the top target of President Donald Trump’s growing trade war.
See also: Intel gains on report of joint venture deal with TSMC
Nvidia has argued that further tightening restrictions will only reinforce China’s determination to make itself independent of US technology and that the clampdown will weaken American companies.
Nvidia’s writedown suggests that the company may miss out on US$14 billion to US$18 billion in revenue for the year, Bloomberg Intelligence analysts Kunjan Sobhani and Oscar Hernandez Tejada said in a note. “If restrictions persist, Nvidia’s data-centre exposure to China could normalize to low- to mid-single digits, similar to early 2024 levels,” they said. That was before production of the H20 ramped up.
The new restrictions follow a National Public Radio report that Trump had backed off from pursuing the H20 controls, in exchange for Nvidia investing in AI data centers. The company just announced that it would build up to US$500 billion worth of AI infrastructure in the US over the next four years, a figure that includes plans that were already underway.
The battle over limits on chip exports has raged on for years. US officials first barred Nvidia and other AI chipmakers from selling their most advanced models to China in October 2022, over concerns that the technology could give Beijing a military edge. Since then, the China controls have ballooned to include an increasingly large set of semiconductor manufacturing tools, as well as a wider range of both processors and high-bandwidth memory chips, which are essential for AI applications.
In addition to capturing a growing set of technologies, the Biden administration also expanded the geographic scope of the AI chip measures — first to some 40 countries that officials worried were providing a backdoor for Chinese firms to access banned chips, and then, in President Joe Biden’s last week, to the entire world. Trump officials have indicated that they want to strengthen and streamline that global framework.
The latest rules for Nvidia are a sign that the Trump administration will stay the course on the US government’s approach to Chinese tech development. They follow earlier sanctions on dozens of Chinese firms that Trump officials allege are aiding Beijing’s military tech efforts.
