Export Controls: 32 Blackwell Ultra B300 GPUs Located in China Telecom Network

Thirty-two Nvidia Blackwell Ultra B300 graphics processing units have been identified within the infrastructure of China Telecom, a state-owned telecommunications provider. The discovery was made by Lava, an Israeli security firm, which noted that these specific chips are subject to strict U.S. export controls that generally prohibit their sale to sanctioned parties. The presence of these high-performance components in a Chinese data center raises immediate questions about the efficacy of current trade restrictions and the mechanisms used to bypass them.

The U.S. government has barred the export of these specific AI data center chips to China, requiring that they be sold only to whitelisted buyers who agree not to transfer them to sanctioned entities. Despite these regulations, the flow of hardware has continued. In 2025, reports indicated that Chinese companies smuggled approximately $1 billion worth of Nvidia AI chips into the country over a three-month period. At that time, the B200 series was the primary target, but distributors had already begun preparing for the release of the superior B300 series.

Nvidia has responded to the latest findings with a mix of acknowledgment and minimization. The company did not deny that the chips were found but characterized the incident as a minor anomaly. An Nvidia spokesperson stated that recent court cases have shown the diversion of controlled U.S. products to be less than half of one percent. The spokesperson described this leakage as a drop in the bucket when compared to the ocean of domestic compute capacity that China already maintains. With tens of millions of GPUs in service in approved countries, the company argued that smuggling remains a non-factor, largely due to governmental enforcement on both sides of the globe.

However, the exact pathway of these specific chips remains unclear. Lava has not confirmed the precise location of the 32 GPUs within China Telecom's network, nor can it explain how they arrived there. The identification was possible through Nvidia's own DCGM Exporter tool. This software is installed by data center operators to monitor server health and performance. Although intended for internal use, the tool can sometimes be accessed from outside the data center network and does not always require password authentication, a security gap that may have facilitated the discovery.

The B300 series is highly sought after because it offers significant performance improvements over its predecessor. Announced in March of last year, the chip provides 50% more memory and FP4 compute capability than the previous-generation B200. Nvidia originally marketed the B300 as a device built for the age of reasoning, specifically referencing sophisticated large language models like DeepSeek R1. The chip was expected to ship by the end of 2025 and quickly became a target for those seeking to expand their AI capabilities.

The smuggling of B300 GPUs has become a focal point in the ongoing battle over AI hardware. In August 2026, nine individuals were indicted in Taiwan for the illegal export of these specific chips to China. This legal action occurred days after Supermicro confirmed it had fired several employees following an internal investigation into smuggling activities. The complex scheme involved a Taiwanese server trading firm named Flying Tiget Tech, which held whitelisted status for GPU purchases, allowing it to acquire hardware that was then diverted to China.

The high demand for Nvidia's silicon has created a market where even older equipment commands premium prices. A100 servers, which are five years old, have seen their prices triple and can now cost up to $82,000. The scarcity of newer chips has also driven some operators to retool gaming GPUs to handle AI workloads where possible. This economic pressure underscores the strategic importance of the Blackwell series and explains why companies continue to risk the complexities of smuggling to acquire them.

The situation highlights a persistent tension between regulatory intent and operational reality. While Nvidia maintains that the percentage of diverted chips is small, the presence of the B300 in China Telecom's network serves as a tangible example of the leaks. The company is currently working on a new tool to help locate its chips, but it has not yet been released. Until such measures are in place, the tracking of export-controlled hardware will likely rely on the same monitoring tools that inadvertently revealed this latest batch.

For now, the 32 chips identified by Lava remain a point of contention. They represent a small fraction of the total hardware in use, yet they symbolize the difficulty of enforcing export controls in a global market hungry for AI power. As Nvidia continues to produce more advanced silicon, the question of how many other units are sitting in unauthorized networks will likely remain open, challenging the notion that current enforcement measures are sufficient to stem the flow.

The discovery does not change the immediate operational status of China Telecom, but it adds another data point to the debate over U.S. tech policy. The company's stance that smuggling is a non-factor may hold statistically, but each confirmed case of B300 diversion serves as a reminder that the physical movement of these chips is far from stopped. As the AI race intensifies, the scrutiny on where these powerful processors end up will only increase.