Nvidia incurs a $5.5 billion penalty as the United States imposes restrictions on chip sales to China.
Nvidia (NVDA.O) announced it will incur $5.5 billion in charges following the U.S. government’s restrictions on exports of its H20 artificial intelligence chip to China, a crucial market for one of its most sought-after semiconductors.
Nvidia’s AI chips have been a primary target of U.S. export regulations, as officials aim to prevent the sale of the most sophisticated chips to China in an effort to maintain a competitive edge in the AI sector. Following the implementation of those constraints, Nvidia commenced the creation of chips that would closely adhere to U.S. restrictions.
Nvidia shares declined almost 6% in after-hours trading.
A representative for the U.S. Commerce Department announced late Tuesday the implementation of new licensing requirements for the export of chips, including Nvidia’s H20, AMD’s MI308, and their counterparts.
The Commerce Department is dedicated to implementing the President’s instruction to protect our national and economic security, stated the spokesperson for the agency responsible for U.S. export restrictions.
AMD did not promptly reply to a request for commentary. Its shares declined by 7% in after-hours trade.
Nvidia’s H20 is its most sophisticated processor offered in China and is pivotal to its strategy for remaining involved in the country’s burgeoning AI sector. Chinese firms such as Tencent (0700.HK), Alibaba (9988.HK), and ByteDance, the parent company of TikTok, have significantly increased their orders for H20 chips in response to surging demand for affordable AI models from the startup DeepSeek, as reported by Reuters in February.
The H20 chip, although not as rapid in training AI models as Nvidia’s chips available outside China, remains competitive with certain models throughout the inference phase, where AI systems provide responses to users. Inference is rapidly emerging as the predominant segment of the AI chip industry. Last month, Nvidia CEO Jensen Huang asserted that the company is strategically positioned to lead in that transition.
Nvidia announced on Tuesday that the U.S. government is limiting H20 shipments to China due to concerns that the processors may be utilized in a supercomputer. Although the H20 possesses inferior computational capabilities compared to other Nvidia chips, its capacity for high-speed connectivity with memory and other computing chips remains substantial.
The memory and networking features may render the H20 advantageous for constructing supercomputers in China, whereas the U.S. has imposed restrictions on the sale of processors for supercomputer applications in China since 2022. On Tuesday, the Institute for Progress, a nonpartisan think tank in Washington, D.C., advocated for the regulation of H20 chips, asserting that Chinese companies were probably already developing such technologies.
One of the purchasers, Tencent, has already implemented H20s in a facility designated for training a substantial model, perhaps in violation of current regulations limiting the use of chips in supercomputers over specific thresholds. The organization stated that DeepSeek’s supercomputer utilized for training its V3 model is probably in violation of the same regulations.
Nvidia said on Tuesday that the U.S. government informed it on April 9 that the H20 chip would require a license to be exported to China and on April 14 it told Nvidia those rules would be in place indefinitely.
The number of licenses the U.S. government may award, if any, remains uncertain.
Nvidia refrained from providing further commentary beyond its submission.
Nvidia stated that the $5.5 billion in charges pertains to H20 items including inventories, purchasing commitments, and associated reserves.
Nvidia announced on Monday its intention to construct AI servers valued at up to $500 billion in the U.S. over the next four years, collaborating with partners like TSMC (2330.TW), aligning with the Trump administration’s initiative for domestic manufacturing.
