China Telecom Network Exposed with 32 Banned Nvidia B300 Chips

A significant breach in the complex web of U.S. export controls has been identified within the infrastructure of a major Chinese state entity. Lava, an Israeli cybersecurity firm, has announced that it has located 32 units of Nvidia’s Blackwell Ultra B300 GPUs inside the network of China Telecom. The presence of these specific chips is notable because they have already been barred from export under recent U.S. regulations, raising immediate questions about the mechanisms used to move high-performance hardware across borders despite official restrictions.

The core of the issue lies in the strict regulations governing the sale of Nvidia’s high-end data center silicon. These chips are typically restricted to whitelisted buyers who contractually agree not to resell them to sanctioned parties, including entities in China. The discovery by Lava challenges the effectiveness of these controls, as the B300 is one of the most powerful graphics processing units currently available for artificial intelligence workloads. It offers substantial performance improvements over its predecessor, the B200, featuring 50% more memory and significantly higher FP4 compute capabilities.

Nvidia has responded to the findings with a mix of acknowledgment and minimization. While the company did not deny the presence of the chips, a spokesperson characterized the incident as a minor anomaly in a vast market. The spokesperson stated that recent court cases have shown that the diversion of controlled U.S. products amounts to less than half of one percent of the total volume. They described this leakage as "a drop in the bucket" compared to the massive amount of domestic compute capacity China already possesses, arguing that with tens of millions of GPUs in service in approved countries, the smuggling remains a non-factor.

The context for this discovery is a history of persistent attempts to circumvent chip bans. In 2025, reports indicated that Chinese companies successfully smuggled approximately $1 billion worth of Nvidia AI chips into the country in just three months. At that time, the B200 was the primary target, but distributors were already positioning themselves to acquire the newer B300 series once it became available. This pattern suggests a well-established pipeline for moving restricted hardware, even as the specific models in question become more advanced and harder to replace.

Technical details from Lava’s investigation reveal how the chips were identified. The firm used Nvidia’s own DCGM Exporter tool, a software utility designed for data center operators to monitor server health and performance. Although intended for internal use, this tool can sometimes be accessed from outside the data center’s network and, in some instances, does not require password authentication. This oversight allowed Lava to pinpoint the location of the B300 units, although they could not confirm the exact physical placement within China Telecom’s broader infrastructure.

The B300 has been a high-stakes commodity since its announcement in March of last year. Nvidia marketed the chip as "built for the age of reasoning," specifically highlighting its suitability for sophisticated large language models. Originally expected to ship by the end of 2025, the GPU quickly became a target for illicit export. In August 2026, the enforcement of these rules tightened, leading to the indictment of nine individuals in Taiwan for the illegal export of B300 GPUs. This legal action followed an investigation by Supermicro, which fired several employees after discovering internal smuggling activities involving a Taiwanese server trading firm with whitelisted status.

The scarcity of these advanced chips has created a ripple effect in the broader hardware market. Demand for Nvidia’s older silicon, such as the A100, has driven prices to triple their original value, with servers now costing up to $82,000 despite being five years old. This price surge underscores the critical role these processors play in AI data centers. As a result, some organizations have resorted to retooling gaming GPUs to handle AI workloads when professional-grade cards are unavailable or too expensive, further complicating the landscape of AI infrastructure development.

Open questions remain regarding how exactly the B300 chips entered China Telecom’s server racks. While Lava has confirmed their presence using standard monitoring tools, the specific route of smuggling is not yet clear. Nvidia has indicated it is working on a dedicated tool to help locate its chips, but this solution has not yet been released. Until then, the gap between official export controls and actual hardware distribution remains a point of tension for policymakers and industry leaders alike.