The collision between Silicon Valley ambition and national security has never looked quite this theatrical. Federal prosecutors just unsealed an indictment against three executives from a billion-dollar American tech company, alleging they orchestrated an elaborate scheme to smuggle AI-powered servers to China using dummy equipment, fake paperwork, and even hair dryers to swap serial numbers. The charges paint a picture of corporate deception that reads more like a spy novel than routine export violations.
At the heart of this case are Yih-Shyan “Wally” Liaw, a 71-year-old U.S. citizen and co-founder of an unnamed publicly traded American manufacturer, along with Taiwanese nationals Ruei-Tsang “Steven” Chang and Ting-Wei “Willy” Sun. According to court documents from the Southern District of New York, the trio allegedly funneled approximately $2.5 billion worth of high-performance servers containing restricted AI graphics processing units to Chinese customers between 2024 and 2025. That figure alone underscores just how lucrative circumventing export controls can be when geopolitical tensions make certain technologies highly coveted.
The technical capabilities at stake here matter tremendously. Modern AI accelerator chips aren’t just fast processors; they represent years of American research investment and engineering refinement. These GPUs can train large language models, process massive datasets for surveillance applications, and power autonomous systems that have dual-use military implications. The U.S. Department of Commerce specifically restricts their export to China and Hong Kong precisely because their computing power could accelerate military AI development or enhance capabilities Washington considers threatening. As MIT Technology Review has extensively documented, the semiconductor rivalry between the United States and China has become central to broader technological competition, with both nations viewing AI chip supremacy as foundational to economic and military strength.
What makes this indictment particularly striking isn’t just the dollar amounts or the technology involved, but the alleged methods used to deceive compliance teams and government inspectors. Prosecutors claim the defendants created thousands of non-functional “dummy” servers—physical replicas that looked convincing but contained no actual computing components. When the company’s own compliance officers or Department of Commerce investigators wanted to verify that servers sold to a Southeast Asian company weren’t being redirected to China, these fake units were supposedly staged at warehouses to pass inspection while the real equipment had already crossed into Chinese territory.
The preparation reportedly involved meticulous attention to cosmetic details. Surveillance footage allegedly captured Sun and an accomplice using hair dryers to carefully remove and reapply manufacturer labels and serial number stickers onto dummy units, then repackaging them in authentic boxes. This wasn’t a hasty cover-up; it was systematic theater designed to withstand scrutiny. The fact that federal authorities obtained video evidence of this preparation work suggests investigators had been monitoring the operation for some time before making arrests.
The scheme’s structure reveals sophisticated understanding of export control vulnerabilities. Rather than attempting direct shipments that would trigger immediate red flags, the defendants allegedly used a company in Southeast Asia as an intermediary. Purchase orders appeared legitimate on paper, with servers assembled in the United States and initially shipped to Taiwan facilities before delivery to this third-party company. Only then were the servers reportedly repackaged in unmarked boxes and forwarded to their actual Chinese destinations. This multi-step process created plausible deniability and made tracking the final endpoints significantly harder for compliance systems designed to flag obvious violations.
Communication security was another layer of the alleged operation. Federal prosecutors note the defendants coordinated using encrypted messaging applications to discuss order quantities, Chinese delivery locations, and strategies for avoiding detection. Wired has reported extensively on how encrypted platforms have become standard tools in both legitimate business operations and illicit schemes, creating challenges for law enforcement while protecting operational security. The indictment suggests these communications provided crucial evidence once investigators gained access.
The timeline is particularly notable. Between late April and mid-May 2025 alone, approximately $510 million worth of U.S.-assembled servers allegedly reached China through this channel. That concentrated burst of activity in just weeks indicates either growing demand, increasing boldness, or perhaps awareness that the operation might soon face scrutiny. John Eisenberg, Assistant Attorney General for National Security, characterized the alleged efforts as involving “false documents, staged dummy servers to mislead inspectors, and convoluted transshipment schemes” specifically designed to obscure that restricted AI technology was heading to China.
The national security implications extend beyond these specific defendants or even this particular company. FBI Assistant Director Roman Rozhavsky emphasized that controlling sensitive AI technology exports remains among the bureau’s highest priorities, framing the investigation within broader counterintelligence efforts. The geopolitical context matters here—Washington has spent years tightening restrictions on advanced semiconductor exports to Beijing, viewing technological advantage in AI as critical to maintaining military and economic competitiveness.
For the American tech industry, this case raises uncomfortable questions about internal compliance systems and corporate culture. How does a publicly traded company with presumably robust export controls allegedly ship billions in restricted technology without detecting the violation internally? Were compliance teams under-resourced, willfully blind, or simply outmaneuvered by executives with access and authority? The indictment suggests the defendants held senior positions—Liaw as co-founder and board member, Chang as general manager—giving them precisely the institutional power needed to manipulate internal processes.
The legal exposure facing the defendants is substantial. Each faces charges including conspiracy to violate the Export Controls Reform Act, which alone carries a maximum 20-year prison term, along with additional counts for conspiracy to smuggle goods and defraud the United States. While maximum sentences rarely materialize in practice, the combination of charges reflects prosecutors’ view of the conduct’s seriousness. Liaw and Sun were arrested and will face proceedings in California, while Chang remains at large, likely in Taiwan.
What happens next will likely influence how other tech companies approach export compliance and how aggressively federal authorities pursue similar investigations. The case demonstrates that substantial resources are being devoted to tracking sophisticated evasion schemes, not just obvious violations. For companies operating in the AI hardware space, the message seems clear: the government is watching, surveillance capabilities are extensive, and the consequences of circumventing controls extend beyond fines to potential decades in prison.
This indictment also arrives as Congress and the executive branch debate how strictly to enforce technology export restrictions without undermining American companies’ competitiveness. Some industry voices argue overly broad controls push international customers toward Chinese alternatives, ultimately weakening U.S. market position. Others counter that preventing adversaries from accessing cutting-edge AI capabilities justifies commercial sacrifices. This case will likely fuel both perspectives—demonstrating both the seriousness of evasion attempts and the challenges of enforcement in global supply chains.
The alleged use of physical dummy servers represents a particularly brazen element that may shape how inspections are conducted going forward. If visual verification can be defeated with convincing replicas, authorities may need to implement more invasive testing protocols or real-time tracking systems. That could increase compliance costs across the industry, even for companies operating entirely within legal bounds.
For now, the case remains in its early stages, with defendants entitled to the presumption of innocence and opportunities to challenge the government’s evidence. But the detailed allegations, supported by surveillance footage and communications records, suggest prosecutors believe they have built a compelling case. Whether this represents an isolated scheme or reveals broader patterns in how export controls are being circumvented remains an open question that will likely drive continued investigation across the tech sector.
