Chinese customs officials have instructed border agents that Nvidia’s H200 artificial intelligence chips are currently not allowed to enter China, according to three sources familiar with the matter. The guidance was issued earlier this week and has already begun affecting shipments, the sources said.
At the same time, Chinese authorities summoned domestic technology firms to meetings on Tuesday, where officials explicitly advised them not to purchase the H200 chips unless absolutely necessary. Two of the sources said companies were warned that approvals would be tightly restricted.
One person briefed on the discussions described the language used by officials as unusually strict, adding that while the move effectively amounts to a ban for now, the policy could still change depending on how the situation develops.
Beijing’s intentions remain uncertain
The H200 — Nvidia’s second most powerful AI processor — has become a major flashpoint in U.S.-China technology relations. Despite strong demand from Chinese companies, it remains unclear whether Beijing intends to permanently block the chip to support domestic semiconductor development, impose temporary limits, or use the restrictions as leverage in broader negotiations with Washington.
The uncertainty is compounded by the chip’s recent approval by the Donald Trump administration for export to China, albeit under specific conditions. In the United States, the decision has drawn criticism from China hawks who argue that advanced AI chips could enhance China’s military capabilities and weaken America’s technological edge.
Sources, who requested anonymity due to the sensitivity of the issue, said Chinese authorities have not provided formal explanations for the directives, nor clarified whether they represent a permanent ban or an interim measure. It also remains unclear whether the restrictions apply to existing H200 orders or only to new purchases.
China’s General Administration of Customs, the Ministry of Industry and Information Technology, and the National Development and Reform Commission did not respond to media inquiries, while Nvidia also declined to comment.
Large orders collide with tightening controls
According to The Information, Chinese officials have indicated that approvals for H200 purchases may only be granted under exceptional circumstances, such as research and development projects conducted in collaboration with universities. Multiple sources said exemptions are currently being discussed for academic and R&D use.
Since 2022, the United States has steadily tightened export controls aimed at slowing China’s progress in advanced AI and semiconductor technology. Last year, Washington initially blocked and later allowed exports of Nvidia’s weaker H20 chip. However, China effectively halted those imports by mid-2024, prompting Nvidia CEO Jensen Huang to state that the company’s AI chip market share in China had fallen to zero.
The H200 represents a far more advanced product, delivering roughly six times the performance of the H20. While Chinese chipmakers have developed alternatives such as Huawei’s Ascend 910C, the H200 is widely viewed as significantly more efficient for training large-scale, cutting-edge AI models.
Chinese technology firms have reportedly placed orders exceeding two million H200 chips, each priced at around $27,000 — far surpassing Nvidia’s available inventory of roughly 700,000 units. Under U.S. export conditions, no more than 50% of total H200 shipments sold to American customers can be allocated to China.
