The United States Department of Commerce has confirmed that a limited number of Nvidia’s advanced H200 artificial intelligence chips have been shipped to customers in China and Hong Kong. The confirmation marks the first official acknowledgment that the highly sought-after processors have crossed the border under the newly established export licensing regime, though the volume remains strictly curtailed.
The shipments represent a delicate balancing act by Washington, which is attempting to maintain technological leverage over Beijing while allowing US semiconductor companies to capture revenue from the massive Chinese market. However, the approved volumes fall far short of Chinese demand, and the US has simultaneously halved its whitelist of approved Asian buyers to crack down on smuggling networks.
Licensed Buyers: ZTE and Maginfra
According to reports from Reuters, the US Commerce Department has approved export licenses for a select group of Chinese companies. Among the named recipients are ZTE Kangxun Telecom, a subsidiary of the telecommunications giant ZTE, and Maginfra, a prominent server manufacturer. A third, unnamed company has also reportedly received approval.
The inclusion of ZTE Kangxun Telecom is particularly notable given ZTE’s complex history with US export controls. The approval suggests that the Commerce Department is willing to grant licenses to major Chinese technology firms provided they can demonstrate that the chips will be used for commercial, non-military applications and are subject to stringent end-use monitoring.
However, the total volume of H200 chips approved for export remains severely restricted. Sources indicate the approved volume is under 200,000 units, less than half the total requested by Chinese buyers. This constraint ensures that while Chinese companies can access the technology for specific projects, they cannot stockpile the chips or deploy them at the scale required to train massive, frontier-level foundation models.
Tightening the Net: Halving the Whitelist
Simultaneously with the approval of the limited H200 shipments, the US government has taken aggressive steps to close loopholes in its export control regime. Nvidia has reportedly halved its whitelist of approved Asian buyers, significantly reducing the number of distributors and resellers authorized to handle its advanced AI chips in the region.
This move is a direct response to the proliferation of smuggling networks that have successfully routed restricted chips into China through third-party countries. Recent investigations have exposed extensive AI chip smuggling networks in a $2.67 billion federal crackdown, highlighting the difficulty of enforcing unilateral export controls in a highly globalized supply chain.
By drastically reducing the number of authorized intermediaries, the US aims to increase visibility and control over the distribution of Nvidia’s hardware. This tightening of the supply chain is intended to ensure that chips exported to countries in Southeast Asia and the Middle East are not subsequently diverted to unauthorized end-users in China.
The Impact on China’s AI Ecosystem
The limited availability of the H200, Nvidia’s most powerful AI processor currently on the market, continues to shape the trajectory of China’s AI industry. While the approved shipments provide a crucial lifeline for specific commercial projects, the overall scarcity of advanced US hardware is forcing Chinese companies to accelerate their transition to domestic alternatives.
The supply constraints have triggered a massive surge in demand for domestic chips, such as the Huawei Ascend 910C. Chinese tech giants, including ByteDance, Tencent, and Alibaba, are increasingly redesigning their AI infrastructure to run on domestic hardware, recognizing that reliance on US technology is a critical vulnerability.
Furthermore, the restrictions are driving architectural innovations within the Chinese semiconductor industry. Companies are exploring alternative approaches, such as near-memory computing and advanced packaging, to extract higher performance from mature manufacturing nodes that are not subject to US sanctions. The recent launch of the DFSX DF1000 AI chip is a prime example of this trend.
A Complex Geopolitical Calculus
The approval of the H200 shipments underscores the complex geopolitical calculus driving US export control policy. The US government is under pressure from the semiconductor industry to allow sales to China, which remains the world’s largest market for commercial chips. A complete embargo would severely impact the revenue of companies like Nvidia, potentially undermining their ability to fund future research and development.
However, Washington is also committed to preventing China from acquiring technology that could enhance its military capabilities or accelerate its development of frontier AI models. The current licensing regime, characterized by limited approvals, strict end-use monitoring, and aggressive enforcement against smuggling, represents an attempt to thread this needle.
As the AI race intensifies, the tension between commercial interests and national security concerns will continue to define the US-China technological relationship. The limited H200 shipments are a temporary compromise, but the long-term trend points toward a deepening bifurcation of the global semiconductor supply chain.