32 Restricted Nvidia B300 Chips Discovered in China Telecom Infrastructure
Security researchers identified banned US-made AI processors within a major Chinese state-owned telecom network, reigniting the debate over the efficacy of export controls.
By Henry Castro ·

Israeli security firm Lava has confirmed the presence of 32 Nvidia Blackwell Ultra B300 graphics processing units inside the network of China Telecom, a state-owned telecommunications giant. These chips are significant because they have been barred from export to China by U.S. regulations, raising immediate questions about how hardware restricted from sale to the country ended up in a domestic data center.
The discovery highlights a persistent loophole in the enforcement of trade restrictions. U.S. export controls stipulate that such high-performance chips can only be sold to whitelisted buyers who agree not to transfer them to sanctioned parties. Despite these rules, the B300 units were located within China Telecom’s infrastructure, a finding that challenges the assumption that export bans are an impenetrable barrier to technology transfer.
This incident is not an isolated event in the history of chip smuggling. In 2025, reports indicated that Chinese entities smuggled approximately $1 billion worth of Nvidia AI chips into the country over a three-month period. While the B200 chips were the primary target during that earlier wave, distributors had already begun preparing for the release of the B300 series, suggesting a sustained and organized effort to acquire the latest silicon.
Nvidia has responded to the news with a mix of acknowledgment and minimization. The company did not deny the findings but characterized the scale of the issue as negligible. A spokesperson stated that in recent court cases, the diversion of controlled U.S. products has amounted to less than half of one percent. Nvidia described this figure as a "drop in the bucket" when compared to the vast amount of domestic compute power China already possesses, arguing that with tens of millions of GPUs in service in approved countries, smuggling remains a non-factor.
The method by which Lava identified the chips offers insight into the security gaps in data center operations. The firm used Nvidia’s own DCGM Exporter tool, a software utility designed for monitoring server health and performance. Although intended for internal use by data center operators, the tool can often be accessed from outside the local network. Notably, it does not always require password authentication, a vulnerability that allowed Lava to detect the specific hardware without physically accessing the server racks.
The motive for acquiring the B300 is clear given its capabilities. Announced in March of last year, the chip offers substantial performance improvements over its predecessor, the B200, including 50% more memory and FP4 compute power. Nvidia had marketed the B300 as being "built for the age of reasoning," specifically referencing advanced large language models like DeepSeek R1. The chip was originally expected to ship by the end of 2025, quickly becoming a highly sought-after asset in the AI sector.
The smuggling of B300 units has already resulted in legal action. In August 2026, nine individuals in Taiwan were indicted for the illegal export of these GPUs to China. This legal development occurred shortly after Supermicro confirmed it had terminated several employees following an internal investigation into smuggling activities. The illicit trade route reportedly involved a Taiwanese server trading firm called Flying Tiget Tech, which held whitelisted status for purchasing Nvidia hardware.
The high demand for Nvidia silicon has created a secondary market where even older generations of chips command premium prices. The company’s A100 servers, which are five years old, have seen their prices triple, with units now costing up to $82,000. This scarcity has led some operators to retool gaming-grade GPUs to handle AI workloads, a stopgap measure that underscores the intense pressure to secure compute resources in a constrained market.
While Nvidia maintains that the volume of smuggled chips is statistically insignificant, the discovery within a state-owned entity like China Telecom suggests that the flow of restricted technology is continuing. The exact location of the 32 chips within China Telecom’s network remains unknown, as does the specific pathway they took to reach the country. Until Nvidia’s forthcoming tool for locating chips is released, the DCGM Exporter tool will likely remain a key mechanism for auditing the presence of controlled hardware in global data centers.
Source: Tom's Hardware · original report published





