Banned Nvidia Chips Found in China Network

A significant tension has emerged in the global semiconductor market after Israeli security firm Lava identified 32 Nvidia Blackwell Ultra B300 GPUs operating within the network of China Telecom. The discovery is notable because U.S. export controls explicitly bar these high-performance chips from leaving the country, raising immediate questions about how they bypassed strict regulatory frameworks to reach a state-owned data center. While Nvidia has not denied the presence of the hardware, the company is pushing back against the narrative that the breach represents a systemic failure, labeling the incident as a negligible anomaly rather than a major leak.

The timeline of these restricted chips entering China reveals a persistent gap between policy and practice. In 2025, reports indicated that Chinese entities smuggled approximately $1 billion worth of Nvidia AI chips into the country over a three-month period. Although the B200 was the primary target at that time, distributors had already begun positioning themselves to sell the newer B300 series once it became available. This pattern suggests that despite ongoing crackdowns, the infrastructure for moving advanced silicon across borders remains robust and active.

Nvidia’s response to the latest findings was swift but measured. A spokesperson told Tom's Hardware that recent court cases have shown the diversion of controlled U.S. products to be less than half of one percent of the total, describing it as a drop in the bucket compared to the vast amount of domestic compute capacity China already possesses. The company emphasized that with tens of millions of GPUs in service in approved countries, this specific instance of smuggling is a non-factor, supported by enforcement actions on both sides of the Pacific.

The methodology used by Lava to locate the chips adds a layer of technical intrigue to the story. The firm identified the B300 units using Nvidia’s own DCGM Exporter tool, a software utility installed by data center operators to monitor server health and performance. Although designed for internal use, this tool can sometimes be accessed from outside the data center’s network and, notably, does not always require password authentication. Nvidia has acknowledged working on a specialized tool to help locate its chips, but that solution has yet to be released.

The B300 GPU is a highly sought-after component, announced in March of last year as a major leap in data center performance. It offers 50% more memory and FP4 compute capabilities than its predecessor, the B200. Nvidia marketed the chip as built for the age of reasoning, citing its utility for sophisticated AI models. Originally expected to ship by the end of 2025, the B300 quickly became a hot commodity, driving intense interest from buyers seeking to enhance their AI infrastructure.

Recent legal actions in Asia underscore the high stakes involved in chip smuggling. In August 2026, nine individuals were indicted in Taiwan for the illegal export of B300 GPUs to China. This indictment occurred just days after Supermicro confirmed it had terminated several employees following an internal investigation into smuggling activities. The complex operations involved a Taiwanese server trading firm, Flying Tiget Tech, which held whitelisted status for purchasing GPUs, highlighting how legitimate channels are sometimes exploited to move restricted hardware.

The scarcity of advanced Nvidia chips has created a ripple effect across the market, driving up prices for older generations of hardware. A100 servers, despite being five years old, have seen their prices triple, with some units costing up to $82,000. The demand for AI compute power is so intense that some operators have resorted to retooling gaming GPUs to handle AI workloads, illustrating the depth of the market’s hunger for Nvidia silicon regardless of its age or official export status.