Nvidia chips banned from China are showing up in Chinese networks

Israeli security firm Lava says it has identified 32 Nvidia Blackwell Ultra B300 processors running inside the network of China Telecom, a state-owned telecommunications provider. The discovery, reported by The Information, raises a direct question: how did these chips reach a Chinese data center when U.S. export controls prohibit their sale to sanctioned parties, including buyers in China? Such processors can only be sold to whitelisted buyers who agree not to re-export them to restricted destinations.

The appearance of these specific chips inside a Chinese state-owned network is significant because it underscores the persistent difficulty of enforcing export restrictions. The B300 is a high-performance data center GPU that is not authorized for sale to China. While the discovery does not explain how the processors were obtained, it does show that controlled hardware continues to reach destinations where it is prohibited.

The path to this latest finding began with a series of escalating restrictions on advanced U.S. technology. In 2025, reports indicated that Chinese companies smuggled $1 billion worth of Nvidia AI chips into the country over three months, at a time when B200 processors were in high demand and distributors were preparing to sell the newer B300 series. More recently, in August 2026, nine people were indicted in Taiwan over the illegal export of B300 GPUs to China. That action came shortly after Supermicro confirmed it had fired several employees following an investigation into smuggling, a process that involved a Taiwanese server trading firm with whitelisted status for GPU purchases.

Nvidia has acknowledged the report but downplayed its importance. In a statement, a company spokesperson said, "Governments, researchers, and media have all found insignificant diversion of controlled US products, less than one half of one percent in recent court cases." The spokesperson added, "With tens of millions of GPUs in service in approved countries, this report again demonstrates that smuggling is a nonfactor, blocked by governments on both sides of the globe." The company did not deny the findings from Lava.

The method of discovery is also noteworthy. Lava identified the chips using Nvidia's own DCGM Exporter tool, software that data center operators install to monitor server health. 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 allowed Lava to locate the processors within China Telecom's server racks. It remains unknown how the chips entered the network or where they are specifically located within it.

The B300 is a highly sought-after component. Announced in March of last year, the data center GPU offers significant performance gains, including 50% more memory and FP4 compute compared to the previous-generation B200. At the time of its announcement, Nvidia described the chip as "built for the age of reasoning," noting its capability for running more sophisticated AI models. The demand for such powerful hardware has led to skyrocketing prices for even older models, with five-year-old A100 servers tripling in price and costing up to $82,000.

The ongoing enforcement of export controls has created a market where some resort to retooling gaming GPUs to handle AI workloads. The presence of restricted processors in a Chinese state-owned network, however, suggests that efforts to circumvent these controls continue. The central question is not just how these 32 chips arrived, but how many others are operating undetected in networks around the world.

The next step for Nvidia is the release of a tool it previously confirmed it was developing, which would allow its chips to be located. The existence of such a tool would provide a way to verify the location of its processors and potentially address the issue of unauthorized diversion. Until then, the company will continue to state that such incidents are a small part of its overall business, while security firms and governments work to track the hardware.