Banned Nvidia Chips Found in Chinese Telecom
Security researchers discovered thirty-two restricted B300 GPUs inside the infrastructure of state-owned China Telecom.
By Rafael Esposito ·

The landscape for global AI infrastructure has shifted again as security experts uncovered a cache of restricted hardware deep within a major Chinese network. Israeli firm Lava identified thirty-two Nvidia Blackwell Ultra B300 chips operating inside the systems of China Telecom. This state-owned telecommunications provider is subject to strict U.S. export controls that explicitly bar the sale of these specific high-performance graphics units. The discovery highlights the persistent difficulty in tracking where advanced computing power ends up once it leaves the hands of approved vendors.
First, the sheer volume of banned hardware is notable. Thirty-two B300 units are not a minor oversight but a significant concentration of power. These chips are among the most capable data center accelerators currently available. They offer substantial performance advantages over the previous B200 generation, including fifty percent more memory and improved FP4 compute capabilities. For any organization building large language models, having access to this hardware can mean a decisive competitive edge in processing speed and efficiency.
Second, the method of detection reveals a gap in standard security practices. Lava located the chips using Nvidia’s own DCGM Exporter tool. This software is designed for internal monitoring of server health and performance. However, it is often accessible from outside the data center network. In some cases, it does not even require password authentication. This technical vulnerability allowed researchers to pinpoint the location of the banned hardware without direct physical access to the server racks.
Third, the situation reflects a broader pattern of enforcement challenges. While U.S. regulations restrict sales to whitelisted buyers who promise not to export to sanctioned parties, smuggling remains a common workaround. In 2025, reports indicated that Chinese entities moved one billion dollars worth of Nvidia AI chips into the country in just three months. The B300 series has become a particular target for these operations. Last month, nine individuals in Taiwan were indicted for illegally exporting these specific GPUs. Supermicro also fired several employees following an internal investigation into similar smuggling activities.
Nvidia has responded to the findings with characteristic confidence. A company spokesperson stated that the diversion of controlled products is less than half of one percent in recent court cases. The company described this figure as a drop in the bucket compared to the massive amount of domestic compute China already possesses. With tens of millions of GPUs in service across approved countries, Nvidia maintains that smuggling remains a non-factor in the overall market balance.
The context of this discovery is driven by intense demand for AI capabilities. The B300 was announced in March of last year and touted as built for the age of reasoning. It supports sophisticated models like DeepSeek R1. Despite the export controls, the chip became hot property quickly. The high stakes have even caused prices for older hardware to skyrocket. A100 servers, which are five years old, have seen their prices triple to reach up to eighty-two thousand dollars. Some buyers are even retooling gaming GPUs to handle AI workloads as a cheaper alternative.
The immediate impact on ordinary users is indirect but tangible. The shortage of high-end chips can influence the cost of cloud services and the development speed of AI features in consumer applications. For now, the focus remains on how regulators and companies will tighten the net around these valuable assets. Nvidia has confirmed it is working on a tool to help locate its chips, but it has not yet been released. Until then, the hunt for banned hardware will likely continue.
The outlook suggests a period of increased scrutiny. Both governments on both sides of the globe are actively trying to block these transfers. Yet the ingenuity of the smuggling networks keeps them in the game. As AI demand grows, the value of these chips will only increase. This makes the enforcement of export controls a critical issue for the future of global technology infrastructure.
Source: Tom's Hardware · original report published





