Banned Silicon: How 32 Top-Tier Nvidia Chips Snuck Into China's State Network

For the millions of professionals relying on high-performance computing and artificial intelligence, the integrity of global supply chains is the invisible backbone of their work. This week, that backbone showed visible cracks when security researchers identified 32 Nvidia Blackwell Ultra B300 graphics processing units operating within the network of China Telecom. This discovery is significant for anyone tracking the flow of technology between the U.S. and China, as these specific chips are currently barred from export by American regulations. Their presence in a state-owned telecommunications provider’s data center raises immediate questions about how restricted hardware is still finding its way into sanctioned territories despite strict government controls.

The identification of these chips was made by Lava, an Israeli security firm, which detected the hardware using Nvidia’s own DCGM Exporter tool. This software is typically installed by data center operators to monitor server health and performance. While designed for internal use, the tool can sometimes be accessed from outside the data center’s network, and it does not always require password authentication. This technical oversight provided a window into the server racks, confirming the physical presence of the B300s even though Lava could not pinpoint their exact location within the company’s broader infrastructure.

The B300 is not an ordinary component; it is one of the most powerful data center GPUs currently available. Announced in March of last year, the chip offers substantial improvements over its predecessor, the B200, including 50% more memory and enhanced FP4 compute capabilities. Nvidia originally marketed the B300 as hardware built for the age of reasoning, specifically citing its suitability for sophisticated large language models like DeepSeek R1. Because of this performance advantage, the chip has become a highly sought-after asset for organizations looking to maintain a competitive edge in AI development, even if they must navigate complex regulatory hurdles to acquire it.

In response to the findings, Nvidia has attempted to downplay the scale of the issue, suggesting that such incidents are rare in the broader context of global chip distribution. A company spokesperson noted that recent court cases have shown that the diversion of controlled U.S. products is often less than half of one percent of the total. The company argued that with tens of millions of GPUs already in service in approved countries, this specific instance is a minor anomaly. Nvidia maintained that smuggling is a non-factor, blocked by governments on both sides of the globe, and that the findings are a drop in the bucket compared to the vast amount of domestic compute power China already possesses.

The arrival of these chips in China follows a pattern of persistent smuggling efforts that have plagued the semiconductor industry. In 2025, reports indicated that Chinese companies managed to move 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 sell the B300 series as soon as it became available. The most recent high-profile case occurred in August of this year, when nine individuals in Taiwan were indicted for the illegal export of B300 GPUs. This legal action came shortly after Supermicro confirmed it had fired several employees following an internal investigation into smuggling operations that utilized a whitelisted server trading firm.

The difficulty in stopping these transfers is compounded by the extreme demand for Nvidia’s silicon. The scarcity of new-generation chips has driven up prices for older models, with five-year-old A100 servers tripling in value and reaching prices as high as $82,000. This economic pressure has led some organizations to retool gaming GPUs for AI workloads as a stopgap solution. As the global race for AI dominance continues, the presence of these 32 B300 chips in China Telecom’s network serves as a reminder that regulatory borders are often tested by the sheer value of the technology crossing them.

Looking ahead, the situation remains fluid. While Nvidia continues to refine its tools for locating its chips, the current methods of detection rely on software that is not always secured against external access. For ordinary people and businesses, the implication is that the availability of top-tier AI hardware is more interconnected and less predictable than official export lists suggest. The ongoing efforts to track these chips will likely continue, as both governments and companies seek to balance the need for technological advancement with the strategic importance of keeping cutting-edge compute power within trusted hands.