The United States has allowed exports of the Nvidia H200 artificial intelligence (AI) chip to China, but China has reportedly prepared measures to restrict imports.

According to Reuters and other outlets, the U.S. Department of Commerce Bureau of Industry and Security (BIS) said in an online bulletin on the 13th (local time) that exports of the H200 and products of the same or lower class to China and Macau would move from a ‘presumption of denial’ to a ‘case-by-case review’ licensing policy. This constitutes an official approval for exports of the H200 to China, which had been prohibited in principle.

Exports are permitted only under stringent conditions. Exporters must demonstrate that domestic supply of the chips in the United States is sufficient and that domestic production will not be disrupted. Export volumes must not exceed 50% of the quantities destined for end users in the United States. Importers must complete BIS customer verification procedures, and the chips must undergo performance testing at an independent institution in the United States. The measures are intended to prevent the chips from being used for military purposes.

President Trump said last month that the U.S. government would allow sales of AI chips to China on the condition that it receives a 25% fee. The decision reflects a judgment that controls on exports of H200-class chips have had little effect in restraining the rapid technological advancement of China. By contrast, China hawks in U.S. politics argued that the move could narrow the U.S.-China technology gap more quickly and be used to strengthen the military power of China.

China, however, appears to have prepared measures that restrict purchases of the H200. The Information, a U.S. IT-focused outlet, reported the same day, citing multiple sources, that China had sent guidance to some technology companies allowing the purchase of the chip only for ‘when necessary’ cases such as university research and development (R&D) institutes.

According to The Information, Chinese authorities initially considered requiring companies purchasing the H200 to buy a certain proportion of domestic AI chips alongside it, but ultimately chose tighter controls. The outlet said the decision appears to prioritize protecting the domestic semiconductor industry, such as Huawei and Cambricon, over the use of advanced chips.

The Information analyzed that Chinese authorities did not clearly set out specific criteria or the scope for what constitutes ‘when necessary’, leaving room for adjustment depending on future changes in U.S.-China relations.

Within China, however, the dominant view is that the import-control measure is intended to spur the technological development of domestic companies and therefore will not be greatly swayed by the course of U.S.-China relations.

According to Reuters, Chinese technology firms had ordered more than 2 million H200 chips priced at about $27,000 each as of last month, exceeding Nvidia stock of 700,000 units. Nvidia Chief Executive Officer Jensen Huang said at CES (Consumer Electronics Show) that production of the H200 is being increased.