Nvidia's B300 Chips Found in Chinese State Network

32 Nvidia Blackwell Ultra B300 graphics processing units have been identified inside the network of China Telecom. The discovery comes from Lava, an Israeli cybersecurity firm that traced the hardware within the state-owned telecommunications provider. These specific chips are currently barred from export to the country by U.S. regulations. The finding raises immediate questions about how restricted technology ended up in a major Chinese data center. It highlights the persistent difficulty in controlling the flow of high-performance AI hardware across borders.

The B300 is a premium data center chip designed for advanced artificial intelligence workloads. It offers significant improvements over its predecessor, the B200. Specifically, the new model provides 50% more memory and 50% more FP4 compute power. Nvidia originally announced the hardware in March 2025, touting it as a chip built for the age of reasoning. The company expected shipments to begin by the end of that year. The chip has since become a highly sought-after asset in the global AI market. Its presence in China underscores the intense demand for top-tier silicon among Chinese firms.

Lava detected the GPUs using Nvidia's own DCGM Exporter tool. This software is intended for data center operators to monitor server health and performance. It is generally designed for internal use but can sometimes be accessed from outside the local network. Oddly, the tool does not always require password authentication. This security gap allowed the researchers to pinpoint the location of the banned chips. It remains unclear exactly where within China Telecom's infrastructure the servers are housed. However, their presence is confirmed by the monitoring data.

U.S. export controls require that these chips be sold only to whitelisted buyers. These buyers must agree not to export the hardware to sanctioned parties, including entities in China. Despite these rules, the flow of Nvidia hardware into the country has not stopped. In 2025, reports indicated that Chinese companies smuggled $1 billion worth of Nvidia AI chips in just three months. The B200 was the primary target then, but distributors were already preparing for the B300 series. The current discovery suggests that the leakage of controlled products remains a significant issue for policymakers.

Nvidia has responded to the report with a mix of acknowledgment and dismissal. A company spokesperson stated that the diversion of controlled products is less than half of one percent. The spokesperson described this figure as a drop in the bucket compared to China's domestic compute capabilities. Nvidia noted that tens of millions of its GPUs are already in service in approved countries. The company insists that smuggling is a non-factor, citing government efforts on both sides of the globe to block unauthorized transfers. Nvidia did not deny the findings but maintained that the impact is statistically insignificant.

The path of the B300 chips into China remains a mystery. It is unknown how they bypassed the standard export channels. However, recent legal actions highlight the complexity of the smuggling routes. In August 2026, nine individuals were indicted in Taiwan for illegally exporting B300 GPUs. This followed an investigation by Supermicro, which fired several employees involved in the scheme. The operation utilized a Taiwanese server trading firm called Flying Tiget Tech. This firm held whitelisted status for GPU purchases, allowing it to act as a conduit for the restricted hardware.

The high value of Nvidia's silicon has driven prices up for older models as well. A100 servers, which are five years old, have seen their prices triple. Some units now cost up to $82,000. This surge in demand has led some operators to retool gaming GPUs for AI workloads where possible. The situation reflects the critical role these chips play in modern data centers. As AI models grow larger and more complex, access to top-tier hardware becomes a strategic advantage. The discovery in China Telecom's network serves as a reminder of the stakes involved in this technological race.

Lava's findings add to a growing body of evidence regarding chip smuggling. The company previously confirmed it was working on a tool to help locate its chips. That tool has yet to be released to the public. Until then, security gaps in monitoring software may continue to reveal the presence of banned hardware. The incident underscores the ongoing tension between U.S. export policy and the global demand for AI computing power. For now, the 32 B300 chips sit within the Chinese network, a tangible result of a complex and costly smuggling operation.

New England readers may note the broader implications for the tech landscape. As AI continues to reshape industries, the availability of hardware determines who can build the most advanced systems. The leak of 32 chips is small in the grand scale of global computing. Yet, it represents a breach in a carefully designed containment strategy. The story illustrates how hard it is to keep cutting-edge technology in specific places when the demand is this high. Nvidia's stance remains that the overall impact is manageable, but the specifics of this case suggest the challenge is far from over.