China Telecom Network Breach Exposes Smuggled Nvidia B300 Chips
A security firm identified 32 export-banned Blackwell Ultra GPUs inside a state-owned Chinese telecom, prompting Nvidia to argue the leakage is statistically negligible.
By Aever Bundji ·
Forty-eight hours after an Israeli security firm disclosed the presence of prohibited hardware in a Chinese state network, the tension between U.S. export controls and global compute demand has reignited. Lava, the security company, announced that it identified 32 Nvidia Blackwell Ultra B300 chips within the infrastructure of China Telecom. These specific processors have been barred from export to China by U.S. regulations, raising immediate questions about how high-value silicon bypassed strict monitoring to land in a data center owned by the state.
The discovery adds a new layer to a persistent market conflict. While official rules dictate that these chips can only be sold to whitelisted buyers who agree not to resell to sanctioned parties, the physical presence of the B300s in China suggests that the enforcement chain is porous. This is not an isolated incident; reports from 2025 indicated that Chinese firms smuggled approximately $1 billion worth of Nvidia AI chips into the country over a three-month period. At that time, the B200 was the primary target, but distributors had already begun positioning themselves for the B300 series as it neared availability.
Nvidia has responded to the news with a strategy of measured dismissal. The company did not deny the findings but characterized the scale of the leak as minor. In a statement, an Nvidia spokesperson noted that recent court cases have found the diversion of controlled U.S. products to be less than half of one percent. The company described this leakage as a "drop in the bucket" when compared to the "ocean of domestic compute" China already possesses. Nvidia emphasized that with tens of millions of GPUs in service in approved countries, smuggling remains a non-factor that is actively blocked by governments on both sides of the Pacific.
The path by which the B300s entered China Telecom’s server racks remains unclear, and Lava has not confirmed the exact location of the chips within the company’s network. However, the identification was achieved using Nvidia’s own DCGM Exporter tool. This software is installed by data center operators to monitor server health and performance. While intended for internal use, the tool can often be accessed from outside the data center’s network. Complicating the security posture further, the tool does not always require password authentication, a vulnerability that has allowed external parties to detect the hardware.
Nvidia previously confirmed it was developing a tool to help locate its chips but has not yet released it. Until then, the DCGM Exporter remains a key vector for detection. The B300 is a highly sought-after asset, having been announced in March 2025. It offers significant performance gains over the previous-generation B200, including 50% more memory and FP4 compute capabilities. At launch, Nvidia marketed the chip as "built for the age of reasoning," specifically citing its suitability for sophisticated AI large language models like DeepSeek R1.
The B300 was originally expected to ship by the end of 2025, and it quickly became a hot commodity. By August 2026, the stakes had risen sharply. Nine individuals were indicted in Taiwan for the illegal export of B300 GPUs to China. This legal action followed days after Supermicro confirmed it had fired several employees following an internal investigation into smuggling. The smuggling operation involved Flying Tiget Tech, a Taiwanese server trading firm that held whitelisted status for GPU purchases, highlighting how legitimate channels can be exploited.
The demand for these advanced chips has also distorted pricing for older hardware. Nvidia’s A100 servers, which are five years old, have seen prices triple due to scarcity. Some units now cost up to $82,000. In response to the ongoing crackdown on new silicon, some operators have resorted to retooling gaming GPUs to handle AI workloads where possible, a move that underscores the critical role Nvidia’s hardware plays in the global AI data center ecosystem.
The next phase of this conflict will likely involve tighter software controls and potentially the release of Nvidia’s new chip-location tool. For now, the presence of 32 B300s in China Telecom’s network serves as a tangible reminder that while export controls are a significant barrier, they are not an impenetrable wall in a market driven by the race for computational power.
Source: Tom's Hardware
