Three universities operating under China's Peoples Liberation Army sought to purchase Nvidia's most powerful artificial intelligence chips through intermediaries, a move that exposes weaknesses in Washington's semiconductor export controls, according to a Bloomberg News investigation published this week.
The Harbin Institute of Technology, the National University of Defense Technology, and the PLA's Air Force Engineering University attempted to acquire Nvidia A100 and H100 processors, which retail for approximately $30,000 each. The transactions were routed through shell companies in Hong Kong and Singapore to bypass restrictions introduced in October 2022.
The Export Control Framework
The Biden administration designed its semiconductor export rules to prevent China from accessing chips capable of training large AI models. Nvidia's A100 and H100 processors sit at the heart of that restriction because they can accelerate machine learning tasks that military strategists and surveillance operations depend on.
The Commerce Department updated its licensing requirements twice since the initial restrictions, yet the attempted purchases demonstrate how Beijing is actively working around the rules. Export control experts note that identifying and blocking every intermediary arrangement stretches the resources of enforcement agencies.
How the Purchases Were Structured
The universities filed procurement requests that listed front companies as end users, Bloomberg reported, citing people familiar with the matter. The actual buyers—entities with clear connections to PLA research programs—remained hidden behind layers of corporate registration in jurisdictions known for permissive business registration rules.
In at least two instances documented by investigators, the chip orders were placed after October 2022, directly violating the spirit of the export ban. Commerce Department officials declined to comment on specific cases but confirmed ongoing investigations into potential violations.
Market Implications for Nvidia
Nvidia shares closed at $875.39 on Thursday, having gained 72 percent over the past twelve months. The company dominates roughly 80 percent of the market for AI training chips, giving it extraordinary pricing power and a backlog of orders stretching into 2025.
Investors have shrugged off export control concerns so far. Wall Street analysts project Nvidia will generate $110 billion in revenue during its current fiscal year, up from $27 billion the year prior. The company has introduced modified chips, such as the H800, designed specifically for the Chinese market to comply with current rules while maintaining performance above regulatory thresholds.
The Secondary Market Problem
The attempted purchases highlight a broader challenge. As long as Chinese entities can access through intermediaries, the export controls function more as a speed bump than a wall. Semiconductor industry consultants estimate that 15 to 20 percent of restricted shipments still reach prohibited users through indirect channels, though precise figures remain difficult to verify.
Semiconductor Supply Chain Tensions
The incident comes as the US government prepares to tighten controls further. Sources familiar with internal deliberations say the Commerce Department is weighing additional restrictions on chips that fall just below current threshold specifications. The goal is to close loopholes that allow slightly less powerful—but still militarily relevant—processors to reach Chinese research institutions.
TSMC, which manufactures Nvidia chips at facilities in Taiwan, has implemented its own screening protocols. However, the contract chipmaker faces limits on what it can independently verify about end customers once products leave its factories.
Beijing's Response and Alternative Strategies
Chinese officials have not publicly addressed the specific university procurement attempts. Beijing has, however, accelerated domestic chip development initiatives, funneling an estimated $150 billion into semiconductor self-sufficiency programs since 2020. The goal is to eliminate reliance on American processors entirely within a decade.
Cambricon Technologies and Huawei's Ascend chip division represent China's most advanced domestic alternatives, though industry benchmarks suggest they remain three to five years behind Nvidia's current generation in AI training performance.
What Happens Next
Commerce Department officials are expected to announce revised export guidelines before the end of the current quarter. Industry observers anticipate the rules will extend restrictions to additional Chinese research institutions and impose stricter documentation requirements for semiconductor purchases in allied nations.
Nvidia is scheduled to report quarterly earnings on May 22. Analysts will be watching for any mention of enforcement actions that might affect future orders or require write-offs of inventory in transit. The company has maintained that it complies fully with all applicable export regulations and terminates relationships with distributors found to violate terms of service.
Semiconductor investors should monitor congressional testimony scheduled for June, when senior Commerce Department officials will face questions from senators about enforcement effectiveness. That hearing could prompt legislative action to allocate additional resources for export control monitoring or introduce criminal penalties for companies that facilitate violations.
