US Government Mandates Licenses for Nvidia H20 Exports

Nvidia Faces New U.S. Export Controls on H20 Chips

In a surprising turn of events, semiconductor giant Nvidia disclosed in a Tuesday filing that the U.S. government has imposed new export controls on its H20 AI chips destined for China. The company has been informed that it will need an indefinite license to proceed with these exports, a move driven by concerns that the H20 chips could potentially contribute to the development of a supercomputer in China.

The financial impact of these new restrictions is substantial. Nvidia anticipates incurring approximately $5.5 billion in related charges during its fiscal first quarter of 2026, which concludes on April 27. Following the announcement, Nvidia's stock experienced a significant decline, dropping about 6% in extended trading sessions.

The H20 chip represents the pinnacle of what Nvidia can currently export to China under existing U.S. export regulations. Interestingly, last week, NPR reported that Nvidia's CEO, Jensen Huang, might have managed to dodge these new restrictions during a dinner at President Donald Trump’s Mar-a-Lago resort. It's rumored that Huang's commitment to invest in U.S.-based AI data centers played a role in these discussions.

Perhaps in a related move, Nvidia announced on Monday its plan to invest hundreds of millions of dollars over the next four years to manufacture some AI chips within the U.S. However, industry observers noted that the company's announcement was somewhat vague on specifics, leaving many questions unanswered.

The call for tighter controls on the H20 chip came from various government officials, spurred by allegations that the chip had been used by the China-based AI startup DeepSeek to train its models, including the R1 "reasoning" model. This model made waves in the U.S. AI market back in January, prompting concerns about its implications.

Nvidia has chosen not to comment on the new export controls or the related developments.