US DOJ Charges Smuggling Ring for Illegal Export of Nvidia AI Chips to China
US federal prosecutors have charged multiple individuals in connection with a smuggling operation that illegally exported at least $160 million worth of Nvidia H100 and H200 AI chips to China, circumventing US export controls. The operation, uncovered as part of Operation Gatekeeper, involved falsifying shipping documents and relabeling chip shipments with a fictional brand to evade detection. One Houston business owner, Alan Hao Hsu, pleaded guilty to smuggling and unlawful export activities, with his company, Hao Global, receiving over $50 million in wire transfers from China to fund the scheme. Two additional suspects, based in New York and Ontario, were also apprehended for their roles in facilitating the illegal shipments, which were coordinated with employees from a Hong Kong-based shipping company and a China-based AI tech firm.
The smuggling ring's activities occurred against the backdrop of shifting US export policies, with the Trump administration recently relaxing some restrictions on older Nvidia Hopper-architecture chips, such as the H100 and H200, while maintaining strict controls on the latest Blackwell chips. The Department of Justice's crackdown highlights ongoing concerns about the transfer of advanced US AI technology to China, particularly for use in high-performance computing and generative AI applications with potential military implications. The case underscores the challenges of enforcing export controls amid evolving technology and international demand for cutting-edge AI hardware.
Sources
Related Stories
US Charges Over Smuggling of Nvidia AI Chips and Supercomputers to China
US federal authorities have charged four individuals—Hon Ning Ho, Brian Curtis Raymond, Cham Li, and Jing Chen—with conspiring to illegally export restricted Nvidia AI chips and supercomputers to China. The indictment alleges that the group used shell companies, falsified invoices, and covert shipping routes through Malaysia and Thailand to bypass US export controls, which have been in place since October 2022 to prevent advanced AI hardware from reaching China. Prosecutors claim the operation involved at least four export attempts, two of which succeeded, resulting in the shipment of approximately 400 Nvidia A100 GPUs to China, while two other attempts involving HPE supercomputers with Nvidia H100 accelerators and 50 Nvidia H200 GPUs were disrupted by law enforcement. The US government asserts that these export controls are critical to limiting China's access to advanced AI technology, which could be used for military modernization and surveillance. The indictment highlights the use of a Florida-based sham real estate company, Janford Realtor LLC, as a front for the purchases, and details how the defendants allegedly doctored customs paperwork to facilitate the smuggling. The case is part of a broader crackdown on efforts to circumvent US technology restrictions, as authorities seek to prevent the transfer of high-performance computing hardware to Chinese entities.
3 months agoNvidia AI GPU Export Controls and Alleged Smuggling to China
U.S. export controls on advanced Nvidia AI GPUs have faced significant challenges as companies with Chinese government ties, such as Megaspeed International, are accused of acting as intermediaries to circumvent restrictions and supply banned chips to China. Investigations by U.S. and Singaporean authorities are underway to determine whether Megaspeed, Nvidia’s largest Southeast Asia customer, facilitated the transfer of restricted AI hardware to China, highlighting the difficulties in enforcing export controls amid globalized supply chains and reseller networks. Nvidia, while denying direct evidence of smuggling, has faced scrutiny over the scale and speed of its shipments and the apparent mismatch between declared data center capacity and imported hardware volumes. Simultaneously, Nvidia is preparing to legally export up to 80,000 H200 AI GPUs to China following a recent U.S. policy shift that allows such shipments under strict conditions, including a 25% revenue share to the U.S. Treasury and recipient vetting. This move marks the first large-scale legal export of high-performance AI chips to China since 2022, as Chinese firms seek access to advanced GPUs that surpass previously allowed variants. The developments underscore the ongoing tension between U.S. efforts to restrict China’s access to cutting-edge AI hardware and the commercial and geopolitical realities of the global semiconductor market.
2 months agoNvidia AI GPU Smuggling and Countermeasures Amid US-China Tech Restrictions
Reports have surfaced alleging that Chinese companies, notably DeepSeek, have illegally obtained thousands of Nvidia Blackwell GPUs by circumventing US export controls through elaborate smuggling operations. These operations reportedly involve setting up fake data centers in Southeast Asia, passing export compliance checks, and then covertly transporting GPU servers into mainland China, where such hardware is restricted. Nvidia has publicly denied knowledge of such smuggling schemes, while unnamed sources claim the use of smaller, easily transportable servers to facilitate these activities. The ongoing demand for advanced AI hardware in China, coupled with US sanctions, has led to persistent efforts to acquire restricted technology through illicit means. In response to concerns about illegal diversion of its AI processors, Nvidia has developed a software-based tracking solution designed to monitor the physical location and health of its GPUs. This technology, which leverages GPU telemetry, aims to help data center operators oversee their fleets and address US political demands to prevent advanced AI chips from reaching restricted markets such as China, North Korea, and Russia. While the tracking capability has not been publicly deployed, it represents a significant step in addressing the challenges posed by export controls and the ongoing technological competition between the US and China in the field of artificial intelligence.
3 months ago