U.S. official confirms 'very few' Nvidia H200 chips shipped to China

During a congressional hearing on Tuesday, Jeffery Kessler, Under Secretary of Commerce for Industry and Security, disclosed that "very few" of Nvidia's H200 artificial intelligence chips have reached China and Hong Kong, describing the total as "a very small quantity of chips." His remarks represent the first official acknowledgment that H200 shipments to China have resumed following months of restrictions. The backdrop to this disclosure is Nvidia's previous policy: since last year, the company excluded any potential Chinese AI chip revenue from its financial forecasts, and in May, CEO Jensen Huang told CNBC investors to "expect nothing" from Chinese sales. An Nvidia representative declined to comment on Kessler's statement. The path to renewed shipments began in December when President Donald Trump said the U.S. government would approve China sales of the H200 chip in exchange for a 25% cut of revenues. Licenses for the H200 were then issued earlier in 2025. Kessler explained that the U.S. government evaluates each company requesting H200 chips on a case-by-case basis, requiring applicants to satisfy national security requirements and submit to inspections to verify compliance. "There are cases where we deny the license applications we receive," Kessler said, signaling that approval is not automatic. Contextually, Nvidia has faced sustained pressure from U.S. export restrictions on most of its products destined for China—a region that represents one of the largest markets for AI development globally. The H200 is an older chip within Nvidia's Hopper generation, less powerful than the newer Blackwell chips that American companies are currently using, a gap that may have made it a more acceptable export item for policymakers. The critical uncertainty ahead is whether China will ultimately approve the import of H200 chips in large volumes. Without access to Nvidia's technology, Chinese firms have been forced to rely on domestic alternatives, which industry observers assess as inferior for AI training tasks—a constraint that could drive demand if Washington permits adequate shipment quantities.

Nvidia's long struggle to access the Chinese AI market has centered on U.S. export restrictions tied to national security concerns and the broader U.S.–China technology competition. For over a year, the company excluded any Chinese revenue from its financial guidance, and CEO Jensen Huang explicitly signaled to investors in May that they should anticipate no material Chinese sales. The Trump administration's December announcement that it would permit H200 exports in exchange for a 25% revenue cut opened a narrow door, and the issuance of licenses earlier in 2025 suggested movement toward resumption. However, Kessler's statement on Tuesday that only "very few" shipments have occurred reflects the continuing tension: while some trade flow has restarted, it remains tightly constrained by case-by-case licensing requirements, national security vetting, and on-site compliance inspections. The H200 itself is an older chip—the Hopper generation—compared to the faster, more powerful Blackwell chips American companies are currently deploying, which limits its strategic value to the U.S. but underscores that any Nvidia shipment to China is politically sensitive.

No comments yet. Be the first to share your thoughts!