New Chip Smuggling Case Exposes Export Control Vulnerabilities

On March 19, 2026, federal agents arrested Yih-Shyan "Wally" Liaw, co-founder and senior vice president of Super Micro Computer, on charges of conspiring to smuggle approximately $2.5 billion worth of Nvidia AI chip-equipped servers to China [1]. Workers allegedly used hair dryers to peel serial number labels off servers and stick them on decoys left behind [2]. The indictment describes a sprawling operation involving shell companies in Southeast Asia, fake purchase orders, unmarked repackaging, and a logistics chain designed to obscure the final destination of some of the most sought-after technology on earth.

The Super Micro case is not an isolated event. It follows December 2025's Operation Gatekeeper, which disrupted a separate $160 million smuggling network [3], and a growing catalog of enforcement actions that together paint a picture of export controls under severe strain. As BIS Assistant Secretary for Export Enforcement David Peters acknowledged: "Simply put, yes, there is chip smuggling, it is going on" [4].

The Super Micro Scheme: Anatomy of a $2.5 Billion Operation

According to the indictment, Liaw and two co-defendants—Ruei-Tsang "Steven" Chang, a sales manager in Super Micro's Taiwan operations, and Ting-Wei "Willy" Sun, a contractor—ran what prosecutors describe as a multi-year conspiracy stretching from 2024 through 2025 [1].

The scheme operated through a specific playbook. Liaw and Chang allegedly directed executives at an unnamed Southeast Asian company to place purchase orders with Super Micro as though the servers were destined for that company's own operations [5]. The servers, containing restricted Nvidia GPUs, were assembled in the United States, shipped to Super Micro's facilities in Taiwan, then forwarded to the Southeast Asian intermediary at a separate location. From there, a logistics company would strip identifying packaging, place the servers in unmarked boxes, and ship them to their actual destination: mainland China [2].

The scale was significant. Between late April and mid-May 2025 alone, servers worth $510 million moved through this pipeline [5]. Each defendant faces three federal charges: conspiracy to violate the Export Controls Reform Act (up to 20 years in prison), conspiracy to smuggle goods (up to 5 years), and conspiracy to defraud the United States (up to 5 years) [1].

Super Micro has been here before. Fortune reported that the company previously faced federal scrutiny over exports to Iran [6].

Operation Gatekeeper: The $160 Million Network

Three months before the Super Micro arrests, in December 2025, the DOJ dismantled a separate smuggling ring through Operation Gatekeeper [3]. This network moved approximately 7,000 Nvidia H100 and H200 GPUs—worth at least $160 million—out of the United States between October 2024 and May 2025 [7].

The key figure, Alan Hao Hsu, a 43-year-old American citizen operating through a Texas-based company called Hao Global LLC, purchased chips from Lenovo's distribution centers in North Carolina [7]. Workers at a New York warehouse then covered Nvidia labels with fake "SANDKYAN" brand names and falsely classified the shipments as "adapters" on export paperwork [7]. The chips traveled from North Carolina to New York, then to Singapore, Hong Kong, and ultimately mainland China [7].

The operation unraveled when BIS intercepted a suspicious shipment in Atlanta in February 2025 [7]. By May, federal agents had intelligence about pallets stored in a New Jersey warehouse. On May 28, in a scene out of a crime thriller, agents conducted surveillance as three trucks arrived to collect the smuggled chips—and moved in to seize over $50 million in Nvidia hardware and cash [3].

Hsu pleaded guilty. Two other defendants, Benlin Yuan (a Canadian citizen and CEO of a Virginia IT services company) and Fanyue "Tom" Gong (a former IBM engineer and Chinese national based in New York), were arrested and charged [3].

How Many Get Caught?

The Center for a New American Security (CNAS) estimated that between 10,000 and several hundred thousand restricted chips were smuggled to China in 2024 alone, with a median estimate of roughly 140,000 units [8]. If accurate, smuggled chips could represent 1–30% of China's inference computing capacity and 1–40% of its training capacity [8].

Online marketplace searches conducted by CNAS researchers found 132 listings for export-controlled chips on Chinese e-commerce platforms, with stock quantities suggesting roughly 100,000 H100 GPUs available domestically as of December 2024 [8]. A CSIS analysis described a "bustling market in Shenzhen" facilitating "sales involving hundreds or thousands of restricted chips" [9].

The Bureau of Industry and Security collected approximately $278 million in criminal and administrative penalties, forfeitures, and restitution in fiscal year 2025 [4]. But the agency's enforcement budget remains a fraction of the profits generated by even a handful of smuggling operations. CNAS found that BIS's annual enforcement budget is less than half the profit from just three documented smuggling cases [8].

Congress approved a 23% increase in BIS's fiscal year 2026 budget [4], and the White House has requested $313 million for the agency—a 64% increase over its current $191 million budget [8]. Whether these resources will match the scale of the problem remains an open question.

The Smuggler's Playbook: Exploiting Systemic Gaps

The cases reveal a consistent set of vulnerabilities in the export control system.

Shell companies and straw purchasers. Both the Super Micro and Operation Gatekeeper schemes relied on intermediary entities that placed orders under false pretenses. In the Super Micro case, an entire Southeast Asian company served as a pass-through [5]. In Operation Gatekeeper, Hao Global LLC was essentially a one-man purchasing front [7].

Transshipment through third countries. Southeast Asia—particularly Singapore, Malaysia, and Thailand—has emerged as the primary transshipment corridor [7]. A 2024 smuggling ring moved $390 million in banned Nvidia GPUs through Malaysia using false pretenses [9]. Chips also transit through Taiwan and Hong Kong before reaching mainland China [5].

False documentation and relabeling. Smugglers systematically falsify shipping manifests, misclassify goods, and physically remove or alter identifying labels [7]. The Super Micro operation used hair dryers to swap serial numbers between real servers and decoys [2]. Operation Gatekeeper defendants relabeled chips with fake brand names [7].

Post-inspection disassembly. One of the more sophisticated methods involves purchasing legitimate data centers in Southeast Asia, passing Nvidia's compliance inspections, then disassembling the entire facility rack by rack and shipping GPU servers across the border into mainland China in suitcases [10].

Exploiting legitimate distribution channels. Both Lenovo and Super Micro—major, publicly traded companies—were used as unwitting participants. Lenovo stated the Operation Gatekeeper purchases were "a standard U.S. domestic sale to a third-party customer who...falsified their identity" [7].

The Price of Forbidden Chips

The economics of chip smuggling are straightforward. Nvidia H100 GPUs that retail for $25,000–$30,000 through authorized channels sell for $40,000–$60,000 or more on gray markets accessible to Chinese buyers [11]. Racks containing eight of the newer B200 GPUs go for $420,000–$490,000 in China, roughly a 50% premium over U.S. prices—meaning sellers can pocket more than $100,000 per transaction [11].

The demand is driven by a clear preference gap. Of 22 major AI models developed entirely in China by 2025, only two used Chinese-made chips [8]. Chinese AI labs strongly prefer American hardware for its "superior performance, higher supply, and more mature software ecosystem" compared to domestic alternatives [8].

The most high-profile allegation involves DeepSeek, the Chinese AI company whose models gained worldwide attention in early 2025. A senior Trump administration official stated that DeepSeek's latest AI model was trained on Nvidia's most advanced Blackwell chips, which would represent a direct violation of export controls [12]. Reports allege the Blackwell GPUs are clustered at a DeepSeek data center in Inner Mongolia [12]. Nvidia responded that it had not "seen any substantiation" of the claims [10].

The Cost to American Industry

Export controls impose costs not just on smugglers but on the U.S. semiconductor industry itself.

The Information Technology and Innovation Foundation (ITIF) estimated that under a full decoupling scenario, U.S. semiconductor firms could lose approximately $77 billion in sales in year one, while competitors in South Korea ($21 billion), the EU ($15 billion), and Taiwan ($14 billion) would absorb the displaced market share [13].

Source: ITIF / Decoupling Risks Report

Data as of Nov 10, 2025CSV

The R&D consequences are direct. U.S. semiconductor companies invested $56.2 billion in R&D in 2024. Under full decoupling, ITIF projects that figure would fall to approximately $43 billion—a 24% reduction, or roughly $14 billion less per year [13]. The employment impact: over 80,000 direct semiconductor industry jobs and nearly 500,000 supporting jobs across the broader economy at risk [13].

These projections represent worst-case scenarios, but even under current partial restrictions, the drag on revenue is real. Qualcomm reportedly faced over $10 billion in lost revenue in 2024 as Chinese customers shifted to alternative suppliers [13]. Micron has signaled it intends to stop supplying server chips to China altogether [13].

The Semiconductor Industry Association has consistently argued that overly broad controls risk undermining the very technological lead they are meant to protect, by starving American firms of the revenue needed to fund next-generation research [14].

China's Response: Accelerated Self-Sufficiency

Whether export controls are slowing or accelerating China's chip ambitions is one of the most contested questions in technology policy.

China has responded to restrictions with a government-backed push toward semiconductor self-sufficiency. Eight of the 19 new global fab construction projects registered in 2024 were located in China [15]. By the end of 2025, Chinese foundries were expected to account for over 25% of global capacity among top mature-node foundries, with China projected to hold the largest installed capacity volume through 2026 [15].

SMIC, China's leading foundry, has expanded aggressively. Its SN1 and SN2 production lines were projected to reach 50,000 wafers per month by end of 2024 [16]. The Financial Times reported that SMIC was on track to mass-produce chips equivalent in performance to the 5nm process node [16]. Huawei has established its own experimental fabrication plant through cooperation with the Shanghai Integrated Circuit Research and Development Center [16].

At the same time, Chinese chip design has shown notable progress. ChangXin Memory Technologies increased China's global memory chip market share from near zero to 5% in five years [9]. Alibaba developed the C930 CPU using RISC-V architecture specifically to circumvent U.S. restrictions on the more common x86 and Arm instruction sets [9]. Researchers at Peking University published work on a 2D transistor that operates 40% faster than TSMC's 3-nanometer devices [9].

CSIS analysts argue these developments suggest the need to shift emphasis from purely restrictive "protect" strategies toward proactive "promote" approaches—investing in domestic research and maintaining the technological edge rather than relying primarily on denial [9].

Legislative and Technological Responses

Multiple legislative efforts are now moving through Congress.

The Stop Stealing Our Chips Act, introduced by Senators Rounds (R-SD) and Warner (D-VA) and backed in the House by Representatives Kean (R-NJ) and Johnson (D-TX), would create a whistleblower incentive program modeled on SEC frameworks. Informants who provide actionable information leading to enforcement penalties would receive 10–30% of resulting fines, with confidentiality protections and anti-retaliation safeguards [17].

The Chip Security Act, introduced by Senator Cotton, would require high-end chip manufacturers to implement technical security measures—potentially including location verification—to detect and prevent smuggling to unauthorized destinations [18].

On the technology front, Michael Kratsios, Director of the White House Office of Science and Technology Policy, confirmed that both software-based and physical tracking solutions are under active discussion [19]. U.S. authorities have already deployed covert tracking devices in select chip shipments to monitor whether they reach China, building legal cases against individuals involved in diversion [20].

BIS may also impose additional compliance requirements on AI server manufacturers, including mandatory chip-level tracking, enhanced end-user verification, and more frequent audits of international shipments [19].

Historical precedent offers some basis for cautious optimism. The Wassenaar Arrangement, which governs dual-use technology exports among 42 participating states, has been adapted multiple times to address emerging smuggling vectors, though its effectiveness has always been constrained by the willingness of member states to enforce its provisions [9]. The Coordinating Committee for Multilateral Export Controls (CoCom), the Cold War-era predecessor, succeeded in restricting Soviet access to certain technologies but could never fully prevent leakage [9].

What Comes Next

The enforcement landscape is shifting rapidly. The Senate Commerce Committee and the House Select Committee on China are both expected to hold hearings on AI chip diversion in coming weeks [19]. The Super Micro and Operation Gatekeeper cases will likely serve as focal points.

But the fundamental tension remains unresolved. Each enforcement action reveals the scale of demand for restricted chips while simultaneously demonstrating the difficulty of controlling the flow of small, high-value goods through global supply chains. BIS has experienced a "steady exodus of employees since the start of 2025," and the Trump administration "forced out two officials focused on efforts to stifle China's technological advances," according to reporting by The Wire China [7].

As independent AI policy analyst Lennart Heim put it after the Operation Gatekeeper arrests: "The question raised by this scheme is, how many more like them are there?" [7]

The answer, based on available evidence, is: more than current enforcement capacity can handle.