32 Blackwell Chips: How Banned Nvidia GPUs Slipped Into China Telecom
Security researchers discovered 32 prohibited Nvidia B300 chips inside a major Chinese state telecom network, prompting Nvidia to call the incident statistically insignificant.
By Samir Nasser ·
For consumers and business leaders watching the AI hardware market, the rules of engagement remain frustratingly porous. Even with strict regulations in place, high-performance compute power continues to find its way into markets where it was explicitly restricted. The latest evidence of this leakage comes from a security audit that revealed a significant quantity of banned chips sitting quietly inside the infrastructure of one of China's largest state-owned telecommunications providers.
Israeli cybersecurity firm Lava has identified 32 Nvidia Blackwell Ultra B300 GPUs within the network of China Telecom. This is a notable find because these specific chips are barred from export to China under current U.S. controls. Typically, these processors are only sold to buyers on an approved list who agree not to resell them to sanctioned parties. The presence of these units in a Chinese data center raises immediate questions about how the supply chain was bypassed.
The discovery is not an isolated incident but part of a broader pattern of chip smuggling. In 2025, reports indicated that Chinese companies moved approximately $1 billion worth of Nvidia AI hardware into the country in just three months. While the B200 model was the primary target then, distributors were already positioning themselves to move the newer B300 series as soon as it became available. The B300, announced in March of last year, offers substantial improvements over its predecessor, including 50% more memory and enhanced FP4 compute capabilities, making it a highly sought-after asset for advanced AI workloads.
Nvidia has responded to the finding with a mix of acknowledgment and minimization. The company did not deny the presence of the chips but argued that the scale of the issue is small compared to the total volume of legitimate sales. A spokesperson stated that recent court cases have shown the diversion of controlled U.S. products to be less than half of one percent. They described this as "a drop in the bucket" relative to the massive domestic compute capacity China already possesses, emphasizing that with tens of millions of GPUs in service in approved countries, smuggling remains a non-factor.
How the chips arrived at China Telecom is currently unclear. Lava confirmed the presence of the hardware using Nvidia’s own DCGM Exporter tool, a software utility designed for data center operators to monitor server health and performance. While intended for internal use, this tool can often be accessed from outside the immediate network and, in some cases, does not require password authentication. This technical oversight allowed the security firm to locate the specific chips, although they have not yet determined their exact physical location within the company’s facilities.
The smuggling of these advanced processors has become a complex logistical operation. In August 2026, nine individuals in Taiwan were indicted for illegally exporting B300 GPUs to China. This legal action followed closely after Supermicro fired several employees following an internal investigation into smuggling activities. The operation reportedly involved a Taiwanese server trading firm called Flying Tiget Tech, which held whitelisted status for purchasing GPUs, allowing it to navigate the export controls more easily than other buyers.
The high demand for Nvidia’s silicon has also distorted the market for older hardware. A100 servers, which are now five years old, have seen their prices triple, with units now fetching up to $82,000. This scarcity has driven some organizations to retool gaming GPUs for AI workloads as a stopgap measure. While Nvidia continues to argue that these losses do not significantly impact the broader market, the persistence of such leaks suggests that the race for AI compute power is still being fought in the shadows.
Open questions remain regarding how widespread this issue is beyond China Telecom. While Nvidia insists that the diversion is negligible, the ability to detect these chips using standard monitoring tools suggests that the infrastructure for tracking and locating banned hardware is becoming more accessible. As the B300 becomes a key component for next-generation AI models, the tension between export controls and market demand is likely to continue testing the boundaries of global tech regulation.
Source: Tom's Hardware
