U.S. uncovers scheme to reroute Nvidia GPUs worth $160 million to China despite export bans

U.S. law enforcement officials announced on Tuesday the dismantling of a smuggling operation linked to China that was attempting to reroute over $160 million worth of export-restricted Nvidia AI chips. In a press release from the U.S. Attorney's Office, it was revealed that two businessmen have been arrested, while a Houston-based company and its owner have already admitted guilt in connection with the broader investigation. This operation, named "Operation Gatekeeper," highlights the U.S. government's intensified efforts to enforce export regulations designed to limit China's access to sophisticated AI technologies, particularly Nvidia's Graphics Processing Units (GPUs). According to U.S. Attorney Nicholas J. Ganjei for the Southern District of Texas, the investigation uncovered attempts to divert advanced AI chips—critical for both military and civilian uses—to parties that pose a potential threat to U.S. national security. Recent documents that have come to light indicate that Alan Hao Hsu, 43, from Missouri City, Texas, along with his firm, Hao Global LLC, pleaded guilty to smuggling and illegal export activities on October 10. Authorities reported that Hsu and his associates had either exported or tried to export Nvidia H100 and H200 GPUs valued at over $160 million between October 2024 and May 2025. While these models are not the latest in Nvidia’s lineup, they still require special licenses for shipment to China due to current regulations. Hsu's operation reportedly involved falsifying shipping documents to disguise the GPUs and obscure their intended destinations, which included China, Hong Kong, and other restricted areas. Investigators traced over $50 million in funds from China that were used to finance Hsu and his company’s illicit activities. Hsu, who is currently out on bail, faces a potential sentence of up to 10 years during his sentencing hearing scheduled for February 18, while Hao Global could face fines amounting to double its illegal earnings, alongside probation. An Nvidia spokesperson commented to CNBC, emphasizing the strict nature of export controls and the scrutiny applied even to older generation products sold in secondary markets. They affirmed their commitment to collaborating with the government and clients to prevent second-hand smuggling. Additionally, U.S. officials have charged Fanyue Gong, 43, a Chinese national living in New York, and Benlin Yuan, 58, a Canadian residing in Ontario, as part of the ongoing investigation. Yuan, who leads a U.S. subsidiary of a Beijing-based IT company, and Gong, who runs a New York tech firm, are accused of conspiring with a Hong Kong logistics firm and a China-based AI company to bypass chip export regulations. Prosecutors allege that Gong employed intermediary purchasers to acquire GPUs under false pretenses, disguising the end customers as being located in the U.S. or other unrestricted countries. U.S. warehouse workers were then instructed to rebrand the shipments under fictitious names and mislabel them as generic components for export to China and Hong Kong. Yuan allegedly coordinated logistics for GPU shipments and developed strategies to obscure Chinese destinations from authorities. If found guilty, Yuan could face a maximum of 20 years for conspiracy to breach the Export Control Reform Act, while Gong may receive a sentence of up to 10 years for smuggling conspiracy. This investigation was conducted with the involvement of the Commerce Department's Bureau of Industry and Security, which enforces U.S. export controls on companies like Nvidia. Amid a series of similar smuggling cases in recent months, lawmakers are working to tighten regulations surrounding U.S. chip exports. Interestingly, the U.S. President recently indicated a willingness to permit Nvidia to ship its H200 chips to "approved customers" in China, contingent on the U.S. receiving a 25% share of the profits. Although the H200 is not the highest-performing chip in Nvidia's range, it would represent the most advanced model available to China, potentially addressing the country's increasing demand for AI computing resources.