Suppliers halt Nvidia H200 output after Chinese customs block shipments

Multiple suppliers that provide components for Nvidia’s high-end H200 AI accelerators paused production after Chinese customs instructed agents to block shipments of the newly approved H200 chips from entering mainland China. The move, combined with near-simultaneous changes to U.S. export policy for advanced AI processors, has put near-term deliveries and revenue to Chinese customers into question.

U.S. export controls were revised on Jan. 13 to move approvals for certain high-end AI chips, including the H200 and a leading AMD model, from a presumption of denial to a case-by-case licensing regime. The Bureau of Industry and Security’s new approach requires U.S. third-party testing of chips’ AI capabilities and includes reporting about a 50 percent volume cap on some sales, with the rules slated to take effect upon publication in the Federal Register. Some earlier reporting referenced a Dec. 8 White House announcement that included a proposed 25 percent government surcharge on exports, a measure that does not appear in subsequent BIS materials.

On Jan. 14, Chinese customs told agents that Nvidia H200 chips were "not permitted" to enter China, and officials reportedly warned domestic IT companies to avoid H200 purchases except under narrow conditions such as research partnerships with universities. Exemptions for R&D and university collaborations are reportedly under discussion, but it is not yet clear whether the customs directive applies only to new orders or also to existing, confirmed shipments.

Industry participants scrambled to respond. On Jan. 16–17 multiple parts suppliers that make components used in H200 systems paused output, halting either final assembly steps or production of precursor parts. Those pauses, described by people familiar with the matter, were implemented after the customs instruction and before the full legal text of the revised U.S. licensing rules appeared in the Federal Register. Suppliers were not named and details on which product lines were affected remain limited.

The commercial mismatch is stark. Chinese firms reportedly placed orders exceeding 2 million H200 chips at roughly $27,000 apiece, a demand level that far outstrips the reported on-hand inventory of about 700,000 units. That gap, together with the customs block and the new U.S. testing and volume limits, could delay shipments even for customers who might qualify under tightened approval criteria.

Analysts and industry lawyers are watching both capitals for signals. Some observers see the customs action as a tactical lever in broader U.S.-China technology tensions, possibly timed ahead of high-level diplomatic meetings expected in April. Others frame the move as a precautionary tightening of controls as both governments recalibrate rules for sensitive AI hardware.

The pause in production exposes downstream suppliers to immediate revenue and scheduling risks and complicates Nvidia’s ability to fulfill high-value orders in the near term. Chinese trade and industry agencies named in coverage have not provided public explanations, and Nvidia did not immediately comment when asked. Key clarifying developments to monitor include the formal publication of the BIS rule in the Federal Register, official guidance from China’s customs and industry ministries on exemptions, and confirmation from suppliers about the scope and duration of production pauses.