Taiwanese prosecutors have indicted nine people — including employees of Nvidia and Super Micro — over the alleged illegal export of AI servers equipped with Nvidia B300 chips to China. The case reveals how export controls on AI hardware are being circumvented through falsified documentation and intermediary shipping routes.
According to prosecutors, the scheme involved falsifying export documents to make 130 B300 AI servers appear as though they were intended for use within Taiwan. In reality, 74 of the servers reached Chinese customers either directly or through intermediaries in Indonesia, Japan, and Hong Kong. Another 56 servers were intercepted by Taiwanese customs authorities before they could leave the country.
The Nvidia B300 servers are subject to U.S. export restrictions that prohibit their sale to China without specific licenses. The restrictions are part of a broader effort to limit China's access to advanced AI computing infrastructure. The indictment shows that enforcement is not limited to individual chips but extends to complete server systems, which contain multiple restricted components and can be shipped through third countries.
For the AI industry, the case demonstrates the growing difficulty of controlling AI infrastructure — not just individual components. Complete servers, which bundle multiple chips, memory, networking, and cooling systems, create new vulnerabilities in the export control framework. The fact that intermediaries in multiple countries were involved highlights the transnational nature of the problem.
The indictment also raises questions about corporate compliance and internal oversight. The involvement of employees at major technology companies suggests that enforcement gaps may exist even within organizations that are subject to strict export control regimes.
