A California computer executive has been arrested in a case that puts the global trade in advanced artificial intelligence hardware under a harsh spotlight. Federal prosecutors allege that Greg Lui, the owner of Earthmade Computer, participated in a conspiracy to export more than $300 million worth of computer servers containing advanced U.S.-manufactured GPUs to China without the required authorization.
According to the U.S. Department of Justice, the alleged Nvidia AI chip smuggling operation ran between 2023 and 2024 and relied on false documentation, overseas intermediaries and shipments routed through Malaysia and Singapore. Prosecutors contend that these third countries were presented as final destinations even though at least some of the Nvidia AI servers were ultimately intended for buyers in China.
The allegations arrive as the United States intensifies enforcement of AI chip export controls designed to prevent China from obtaining the most powerful American computing technology. Advanced Nvidia GPUs can train large AI models, operate massive data centers and support high-performance computing at a scale with potential commercial, military and intelligence applications.
The case remains pending. An indictment contains allegations, not proof, and Lui is presumed innocent unless and until the government establishes guilt beyond a reasonable doubt in court.
What the Justice Department alleges against Greg Lui
The government alleges that Lui and unnamed or separately identified co-conspirators arranged the sale and export of high-performance servers equipped with export-controlled GPUs. Earthmade Computer, the California business owned by Lui, is alleged to have served as a central supplier in the operation.
Prosecutors say the Greg Lui China smuggling case involved more than a few isolated shipments. The total value of the computer servers allegedly exported between 2023 and 2024 exceeded $300 million, placing the matter among the most consequential publicly reported enforcement actions involving restricted AI computing equipment.
The indictment reportedly charges Lui with federal offenses related to conspiracy, violations of U.S. export controls, smuggling and financial transactions connected to the alleged scheme. The precise elements of each charge must be proven separately. Prosecutors generally must establish not only that restricted equipment left the United States without the necessary license, but also that a defendant knowingly participated in the unlawful conduct required by the relevant statute.
The Justice Department maintains that the shipment records did not accurately disclose the intended destination of the hardware. Instead, the alleged conspirators used documents identifying companies in Malaysia and Singapore as purchasers or recipients while planning to redirect equipment to China.
How the alleged Nvidia AI chip smuggling scheme worked
Export-control evasion often depends on creating distance between a restricted U.S. product and its true end user. In the Earthmade Computer Nvidia case, prosecutors allege that third-country businesses and shipping routes provided that distance.
The alleged process can be summarized in several stages:
- Earthmade allegedly obtained or sold servers containing advanced U.S.-manufactured GPUs subject to export restrictions.
- Shipping and commercial records allegedly identified Malaysia or Singapore as the destination, customer or end-use location.
- Prosecutors say false statements and documentation concealed the hardware’s intended connection to China.
- The equipment was allegedly shipped through overseas intermediaries before being re-exported or redirected to China-based buyers.
- Payments from foreign shipping companies allegedly flowed back to Earthmade, creating a financial trail tied to the transactions.
Malaysia and Singapore are legitimate international technology and logistics hubs, and routing a shipment through either country is not inherently unlawful. The government’s theory is more specific: prosecutors allege that entities in those jurisdictions were used to disguise the actual destination and evade U.S. AI chip restrictions on China.
This distinction is important. Export compliance does not end when a product reaches the first overseas consignee. U.S. rules can require exporters to examine the end user, ultimate consignee, intended use and risk of diversion. A transaction may raise concerns when a newly formed intermediary orders unusually large volumes of advanced hardware, provides vague end-use information or operates from a location inconsistent with the computing capacity being purchased.
The alleged $176 million payment trail
Financial records form a major part of the government’s allegations. Prosecutors report that Earthmade received approximately $176 million from Malaysia-based shipping companies. Those transfers allegedly corresponded with purchases and movements of advanced servers.
The payment figure does not, by itself, prove illegal export activity. International hardware transactions routinely involve distributors, freight companies and overseas customers. Prosecutors will need to connect the money to restricted shipments and prove the knowledge and intent required for the charged offenses.
Nevertheless, payment patterns can help investigators reconstruct an alleged export network. Bank records may identify who funded purchases, which businesses acted as intermediaries and whether the commercial explanation matches shipping documents. A logistics company sending substantial payments to a computer supplier, for example, may invite scrutiny when the same transactions involve controlled GPUs and uncertain end users.
The reported $176 million in payments also illustrates why enforcement increasingly combines export-control investigations with financial analysis. Authorities can compare invoices, wire transfers, freight records, serial numbers, customer communications and end-user certifications to determine whether a stated sale was consistent with the hardware’s actual route.
The 27-server shipment cited by prosecutors
One example in the indictment involves 27 servers valued at approximately $7.6 million. Prosecutors allege that the servers were represented as part of an overseas transaction but were ultimately sent to a buyer based in China.
This shipment offers a more concrete view of the broader $300 million Nvidia chips China allegation. At an average value exceeding $280,000 per server, the equipment was not ordinary office technology. Enterprise AI servers can contain multiple high-end GPUs, specialized networking components, large memory capacity and power systems designed for computationally intensive workloads.
The government is expected to rely on records showing how the 27 servers were ordered, paid for, exported and transferred. The defense may challenge whether Lui knew their ultimate destination, whether the products were controlled under the applicable regulations at the time and whether documents or communications support the prosecution’s interpretation.
Why Nvidia H100-class GPUs face export controls
Modern AI development depends heavily on parallel processing. Advanced GPUs can perform enormous numbers of mathematical operations simultaneously, making them particularly effective for training and running large language models, image generators, scientific simulations and other demanding systems.
Nvidia H100-class technology became central to the export-control debate because it delivers high computational performance and can be linked into large clusters. The concern behind Nvidia GPU export controls is not a single chip operating alone. Thousands of accelerators connected through high-speed networks can create an AI supercomputer capable of training sophisticated models or processing vast datasets.
Advanced AI chips can support beneficial work in medicine, climate science, manufacturing and cybersecurity. The same underlying computing capacity may also be used for military modeling, intelligence analysis, autonomous systems, cyber operations and surveillance. U.S. semiconductor export controls seek to limit certain combinations of chip performance, interconnection speed and aggregate computing capability from reaching restricted destinations or end users.
The rules have evolved as manufacturers developed compliant products and computing architectures changed. Exporters therefore must evaluate the exact chip, server configuration, destination and end user rather than assuming every Nvidia product is treated identically. The U.S. Commerce Department’s Bureau of Industry and Security administers the primary export-control framework and provides regulations, licensing guidance and restricted-party resources.
Why Malaysia and Singapore matter to enforcement
The alleged Malaysia Singapore Nvidia shipments reflect a broader challenge for regulators: restricted technology may be diverted through countries that maintain extensive commercial relationships with both the United States and China.
Neither country should be treated as a presumptively illicit destination. Both are major participants in global electronics, cloud computing and semiconductor supply chains. Their importance, however, means that exporters must be alert to transactions in which a local buyer appears to lack the facilities, power capacity, technical staff or business purpose needed to use a large quantity of AI hardware.
Authorities are increasingly focused on beneficial ownership, transshipment behavior and data-center capacity. A shipment may appear compliant on paper while the actual customer controls the intermediary from another jurisdiction. Likewise, servers can be physically delivered to one country but remotely operated for a restricted party or quickly forwarded elsewhere.
This is why effective AI hardware export restrictions involve more than checking an address. Suppliers may need to verify corporate ownership, request detailed end-use statements, screen related parties, examine payment sources and investigate abrupt changes in shipping instructions.
A test of broader U.S. AI chip restrictions on China
The arrest is part of a wider U.S. effort to prevent advanced AI chips China-based organizations cannot lawfully purchase directly from obtaining through intermediaries. Enforcement has expanded alongside the rules because restrictions have limited value if buyers can simply route the same products through another market.
The potential rewards for diversion are substantial. Demand for Nvidia AI computing hardware has outpaced supply at various points, while access restrictions can create steep price differences between authorized and prohibited markets. These conditions may encourage an AI chip black market involving resellers, shell companies, false end-user statements and complex logistics chains.
At the same time, regulators face a difficult balance. Overly broad controls can disrupt legitimate cloud providers, universities, startups and data centers outside China. Narrow rules may leave pathways for sophisticated buyers to combine lower-performance products or obtain restricted systems through third parties.
The alleged AI chip smuggling operation highlights how U.S.-China AI chip exports have become both a national-security issue and a compliance challenge for the private sector. Manufacturers, distributors, server integrators, freight forwarders and financial institutions can all hold information needed to identify suspicious transactions.
What the case means for technology suppliers
Companies selling advanced computing systems should view the allegations as a reminder that customer screening alone may be insufficient. A distributor that is not on a restricted-party list can still present diversion risk based on its ownership, location, purchasing history or intended use.
Practical safeguards include:
- Classifying every GPU and complete server under the applicable export-control rules.
- Confirming whether a license is required for the destination, end user or end use.
- Verifying beneficial owners, payment sources and ultimate consignees.
- Comparing purchase volumes with the customer’s facilities and stated business model.
- Prohibiting unauthorized re-export and remote access in contractual terms.
- Escalating altered shipping instructions, inconsistent documents and third-party payments.
- Maintaining invoices, communications, serial numbers and logistics records for audits.
No compliance system can eliminate every risk, but documented due diligence can expose inconsistencies before controlled hardware leaves the United States. It can also help a legitimate exporter demonstrate that it took reasonable steps to understand a transaction.
What happens next in the Greg Lui Nvidia chips case
Following the arrest, the prosecution must prove the charged conduct through admissible evidence. Pretrial proceedings may address detention, discovery, digital records, overseas evidence and challenges to the indictment. If the case reaches trial, jurors would determine whether the government established each element beyond a reasonable doubt.
As of October 2026, descriptions of the alleged conduct should not be read as a finding that Lui or any other defendant committed a crime. Lui is entitled to contest the government’s evidence and present defenses. An indictment represents the prosecution’s accusations and does not carry a presumption of guilt.
Whatever the eventual outcome, the U.S. Department of Justice Nvidia case demonstrates that advanced computing supply chains are now a major enforcement priority. The government is signaling that it will follow both servers and payments across borders when it suspects restricted AI technology is being diverted to China.
Frequently asked questions
Who is Greg Lui?
Greg Lui is identified by federal prosecutors as the California owner of Earthmade Computer. He has been arrested and accused of participating in a scheme to export controlled computer servers to China. The accusations have not been proven, and he is presumed innocent unless proven guilty.
How much Nvidia equipment was allegedly exported?
The Justice Department alleges that the scheme involved more than $300 million in computer servers containing advanced U.S.-manufactured GPUs. Prosecutors also report that Earthmade received approximately $176 million from Malaysia-based shipping companies.
Why were the servers allegedly routed through Malaysia and Singapore?
Prosecutors contend that businesses and shipping routes in Malaysia and Singapore were used to make those countries appear to be legitimate destinations for the equipment. The government alleges that restricted servers were instead re-exported or redirected to buyers in China.
Are all Nvidia GPUs prohibited from export to China?
No. Export treatment depends on a product’s technical specifications, destination, end user, end use and the regulations in effect at the time. Certain advanced data-center GPUs and systems require authorization, while other products may fall outside the same restrictions.
What can advanced Nvidia AI servers be used for?
They can train and operate large AI models, perform scientific simulations, analyze extensive datasets and power high-performance computing clusters. These capabilities have valuable civilian applications but can also support military, intelligence, surveillance and cybersecurity activities, which is why the most advanced systems face heightened controls.
