A top U.S. Commerce Department official confirmed Tuesday that "very few" Nvidia H200 AI chips have been shipped to China under licenses issued earlier this year, signaling a limited but notable reopening of a market Nvidia had told investors to discount entirely.
The United States government has confirmed that a limited number of Nvidia H200 artificial intelligence chips have been shipped to China and Hong Kong under licenses issued earlier this year, a disclosure that signals the partial reopening of a market that Nvidia's own leadership had told investors to write off entirely.
"The bottom line is very few shipments against licenses for H200s and equivalents have taken place. It's a very small quantity of chips," Under Secretary of Commerce for Industry and Security Jeffery Kessler said at a congressional hearing on Tuesday.
The remarks represent the clearest official acknowledgment yet that H200 exports to China have resumed, even if on a limited scale, following months of near-total restriction stemming from the ongoing technology and trade dispute between Washington and Beijing.
Nvidia's CEO Jensen Huang told CNBC in May that he had instructed investors to "expect nothing" from Chinese sales, and the company has excluded any potential Chinese AI chip revenue from its financial forecasts since last year. An Nvidia representative declined to comment on Kessler's testimony.
The path to resumed shipments traces back to December, when President Donald Trump indicated the administration would approve H200 sales to China in exchange for a 25% cut. Licenses were subsequently issued earlier this year, though Kessler's comments make clear that actual shipments remain far below what those licenses could theoretically permit.
The Commerce Department's approach to approvals is deliberate and selective. Kessler said the government evaluates each applicant on a case-by-case basis, requiring companies to demonstrate compliance with national security standards and submit to inspections. "There are cases where we deny the license applications we receive," he said.
The H200 sits in Nvidia's older Hopper chip generation. American companies have already moved on to the faster Blackwell architecture, meaning the chips being shipped to China represent hardware that is no longer at the cutting edge of Nvidia's lineup.
Whether China will ultimately permit large-scale imports of the H200 remains an open question. Without access to Nvidia's products, Chinese AI developers must rely on domestically produced alternatives, which are widely regarded as less capable for the computationally intensive task of training large AI models.
China is among the largest markets globally for AI development, and Nvidia has long sought to maintain a commercial foothold there. The export restrictions imposed over recent years have effectively severed much of that relationship, making even a trickle of licensed shipments a notable shift in the trade landscape.
The modest volume of chips that have reached China so far suggests the licensing framework, while active, has not yet translated into the kind of commercial volume that would materially affect Nvidia's revenue projections. Whether that changes will depend on both Washington's willingness to expand approvals and Beijing's receptiveness to the imports.
TechEchelon Staff bylines are produced collectively by the newsroom for short, breaking, and wire-style coverage. Longer-form reporting is published under the responsible reporter's name.
More from the Staff →Grassroots opposition to AI data center construction has more than doubled in a single quarter, with protesters blocking or delaying at least 75 U.S. projects valued at $130 billion as communities push back against rising energy costs, water concerns, and environmental impacts.
NHTSA administrator Jonathan Morrison has issued a directive demanding that autonomous vehicle developers fix their systems' failure to detect and respond to emergency scenes, with a deadline for solutions set for the end of July.
Goldman Sachs initiated coverage on Hong Kong-listed Zhipu with a price target of 1,880 Hong Kong dollars while naming DeepSeek and ByteDance as its other top Chinese AI model picks, citing near-frontier performance and surging enterprise adoption.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.