A major enforcement action in Taiwan has exposed an alleged scheme to smuggle Nvidia H200 AI servers to China, underscoring the persistent and lucrative black market that has emerged around US export controls on advanced semiconductors. Taiwan’s Keelung District Prosecutors’ Office recently executed search warrants at 12 locations across the island in the island’s first formal crackdown on illicit AI semiconductor exports to China. Three suspects are accused of forging shipping documents to smuggle Supermicro servers containing Nvidia AI chips into China, Hong Kong, and Macau. The Taiwan action is separate from, but closely connected to, the much larger US federal prosecution unsealed in March, in which the US Department of Justice charged Supermicro co-founder Yih-Shyan “Wally” Liaw and two others with conspiring to smuggle approximately US$2.5 billion worth of Nvidia-equipped servers to China through shell companies in Southeast Asia. Liaw has pleaded not guilty.
The Scale of the Operation
The H200, Nvidia’s current-generation AI accelerator, has been licensed for export to China but no units have been delivered, the license exists on paper while enforcement actions and compliance reviews have stalled actual shipments. The chip’s performance specifications make it one of the most sought-after items on the illicit technology market, commanding significant premiums over its official list price when acquired through gray-market channels. Supermicro, headquartered in San Jose, California, is one of the world’s largest server manufacturers and a major distribution partner for Nvidia. The company has faced regulatory scrutiny in the past, including a 2023 investigation into accounting irregularities. The latest raids add to a pattern of compliance challenges that have dogged the company’s operations in Asia and raise questions about the adequacy of the due diligence processes Supermicro applies to its distribution network.
Jensen Huang’s Response
Nvidia CEO Jensen Huang responded publicly to the raids by urging Supermicro to “tighten compliance.” Huang emphasized that Nvidia maintains strict internal controls over its products and that any violations of export regulations are ultimately the responsibility of the channel partners who handle distribution. The statement reflects Nvidia’s difficult position: the company is legally barred from selling its most advanced chips to China, yet it has limited ability to control what happens to its products once they enter the distribution chain.
Nvidia’s compliance obligations extend beyond simply refusing to sell restricted chips directly. The company is also expected to implement end-use monitoring programs and to take action against distributors who are found to have violated export controls. Huang’s public statement can be read as both a genuine call for better compliance and a preemptive distancing of Nvidia from liability in the event that regulators pursue further action.
A Widening Enforcement Gap
The Taiwan bust is part of a broader pattern of illicit technology transfers that enforcement agencies have struggled to contain. Earlier this year, a separate federal crackdown exposed AI chip smuggling networks valued at $2.67 billion, demonstrating that the demand for restricted chips in China remains intense despite, or perhaps because of, the tightening of controls.
The progression of US restrictions has been relentless: the H100 was banned in 2022, the H800 and A800 in 2023, and the H20 in 2025. Each successive ban has pushed Chinese buyers to seek more creative workarounds, while simultaneously accelerating domestic chip development efforts. The US Senate has also sought to extend restrictions further, with legislation introduced to block AI chipmaking equipment sales to China. As long as a significant performance gap exists between permitted and restricted chips, the incentive to circumvent export controls will remain powerful, presenting an ongoing challenge for both regulators and compliant industry participants.
The Taiwan raids also highlight the geographic complexity of the enforcement challenge. Taiwan sits at the center of the global semiconductor supply chain, and its role as a manufacturing and logistics hub makes it both a critical partner for US export control enforcement and a natural transit point for those seeking to circumvent it.
The Cost of Compliance for the Industry
For legitimate players in the AI server supply chain, the enforcement environment is creating significant compliance costs. Companies must invest in enhanced due diligence processes, end-use verification programs, and legal resources to navigate an increasingly complex regulatory landscape. Supermicro’s experience illustrates the reputational and legal risks that arise when these processes fail, whether through negligence or deliberate circumvention by actors within the distribution network.
The broader industry implication is that the era of relatively frictionless global technology distribution is coming to an end. The AI chip supply chain is being bifurcated into a Western-aligned segment, subject to strict export controls, and a China-facing segment that must operate under a different set of rules. Companies that straddle both markets face growing pressure to choose sides or invest heavily in the compliance infrastructure needed to operate in both.
For Nvidia, the Supermicro situation is a reminder that its commercial interests and compliance obligations are in permanent tension. Every H200 server that reaches China through illicit channels represents both a lost legitimate sale and a potential regulatory liability. Managing that tension, while continuing to grow in the markets where it is permitted to operate, will remain one of the defining challenges of Nvidia’s business for the foreseeable future.