Banned Nvidia Chips Found in Chinese Telecom Network
Security researchers found 32 high-end Nvidia GPUs in China Telecom, prompting the chipmaker to argue that such leaks are minor compared to the total volume of hardware in use.
By Pavel Bell ·

U.S. export controls aim to limit the flow of advanced technology to specific countries. Despite these rules, 32 banned Nvidia GPUs were recently found inside the network of a major Chinese state-owned telecom provider.
Lava, an Israeli security firm, identified the Nvidia Blackwell Ultra B300 chips within China Telecom. This state-owned company provides telecommunications services across China. The B300 is a high-performance chip that has been barred from export under current U.S. regulations. These controls typically restrict sales to whitelisted buyers who agree not to pass the hardware to sanctioned parties. The discovery raises questions about how the chips entered a Chinese data center despite these safeguards.
Previous reports indicate that export bans have not always stopped the flow of hardware. In 2025, Chinese companies reportedly smuggled $1 billion worth of Nvidia AI chips into the country in just three months. At that time, the B200 model was the most sought-after, but distributors were already preparing to move B300-series chips as soon as they became available. The current finding suggests that the issue of controlled chips reaching China remains a persistent challenge for regulators.
Nvidia has responded to the news by downplaying the impact. A spokesperson told Tom's Hardware that the diversion of controlled products is negligible. The company noted that recent court cases show less than half of one percent of products are diverted. Nvidia compared this figure to the vast amount of domestic compute power China already possesses. The spokesperson stated that with tens of millions of GPUs in service in approved countries, smuggling remains a non-factor. The company believes governments on both sides of the globe are effectively blocking the flow.
The exact method of entry for these specific chips is unknown. Lava could not confirm the precise location of the hardware within the company's network. However, the firm identified the chips using Nvidia's DCGM Exporter tool. This software allows data center operators to monitor server health and performance. It is designed for internal use but can often be accessed from outside the data center network. Notably, the tool does not always require password authentication, which may have aided the discovery. Nvidia has confirmed it is working on a tool to locate its chips but has not released it yet.
The Blackwell Ultra B300 is highly desirable because it offers significant performance gains. Announced in March 2025, the chip provides 50% more memory and FP4 compute than its predecessor, the B200. Nvidia described the hardware as built for the age of reasoning, citing its suitability for sophisticated AI models like DeepSeek R1. The chip was expected to ship by the end of 2025 and quickly became a hot commodity. In August 2026, nine people were indicted in Taiwan for illegally exporting B300 GPUs to China. This followed an investigation by Supermicro, which fired several employees linked to smuggling operations involving a whitelisted trading firm.
The high demand for Nvidia chips has driven up prices for older models. A100 servers, which are five years old, have seen prices triple and can now cost up to $82,000. The crackdown on smuggling has led some buyers to retool gaming GPUs for AI workloads where possible. The next step for Nvidia is to release its new tracking tool. Meanwhile, regulators continue to monitor the flow of hardware to ensure compliance with export laws.
This discovery highlights the ongoing tension between trade restrictions and the global demand for AI infrastructure. The presence of these chips in a state-owned network underscores the difficulty of enforcing export controls in a complex supply chain. As Nvidia works to improve its tracking capabilities, the industry watches to see if these measures will reduce future incidents of illegal diversion.
Source: Tom's Hardware · original report published





