A ChosunBiz report relaying Reuters’ Nov. 5, 2026, account said about 600,000 Nvidia H200 AI accelerators were being held in Hong Kong while awaiting shipment to mainland China. The estimate puts the spotlight on two separate hurdles: U.S. export clearance and China’s approval of large-scale imports.
Why export clearance did not settle the import question
The report said U.S. export rules had been adjusted to permit sales to approved Chinese customers, while Chinese authorities had not approved large-scale H200 imports. Limited quantities had reportedly entered mainland China through authorized or indirect channels.
ChosunBiz described China’s policy as allowing limited U.S. chips for AI training while favoring domestic chips for inference. Training is the process of building or updating a model; inference is when a trained model generates an answer or prediction. The distinction helps explain why some H200s could enter while large-scale imports remained restricted.
A U.S. export clearance and Chinese import approval govern different steps in a shipment. Permission on one side does not by itself move the hardware through the other.
A sale can clear one government’s export rules and still face a separate import decision at its destination.
The reported limited arrivals are distinct from the Hong Kong estimate. NeoTeo previously covered reports of H200 shipments to ByteDance and Tencent.
What the May clearance report said
On May 14, 2026, Reuters reported that around 10 Chinese firms had U.S. clearance to buy H200 chips, but no deliveries had occurred as of that report. That was an earlier snapshot, months before the November stockpile report.
What Lee Na-yeon said about Nvidia’s exposure
In a Korean-language analysis published Nov. 5, host Lee Na-yeon said Nvidia’s then-current quarterly revenue guidance of around $108 billion did not rely on data-center chip sales to China. She also cited a $400 million cost related to H200 inventory and pre-commitments in the preceding fiscal half-year.
