Nvidia Blackwell Ultra B300 GPUs Discovered in China Telecom Network

Lava, an Israeli cybersecurity firm, announced this week that it has identified 32 Nvidia Blackwell Ultra B300 graphics processing units operating within the network of China Telecom. The state-owned telecommunications provider serves a domestic audience, making the presence of these specific chips a notable development in the ongoing battle over high-performance hardware. The discovery highlights the continued difficulty in restricting the flow of advanced American silicon to Chinese data centers, despite strict export controls.

The B300 series has been barred from export to China under U.S. regulations, which means these chips should only be sold to approved buyers who agree not to pass them on to sanctioned entities. The presence of the hardware in a Chinese data center raises questions about how the units bypassed these checks. This incident follows a broader trend of smuggling, with reports from 2025 indicating that Chinese companies moved approximately $1 billion worth of Nvidia AI chips into the country in just three months. At that time, the B200 model was the primary target, though distributors were already preparing for the B300 release.

Nvidia responded to the report by acknowledging the findings but downplaying their significance. A company spokesperson stated that governments, researchers, and media have found only insignificant diversion of controlled products, citing recent court cases where the diversion rate was less than half of one percent. The spokesperson described this as a drop in the bucket compared to the vast amount of domestic computing power China already possesses. Nvidia emphasized that with tens of millions of GPUs in service in approved countries, smuggling remains a non-factor that is blocked by governments on both sides of the globe.

The exact method by which the B300 chips entered China Telecom's server racks remains unknown. Lava has not been able to confirm the precise location of the units within the company's network, but it confirmed their presence using Nvidia's DCGM Exporter tool. This software is designed for data center operators to monitor server health and performance. While intended for internal use, the tool is often accessible beyond the immediate data center network and does not always require password authentication, a security gap that has allowed for external detection.

The desire for the Blackwell Ultra B300 is driven by its substantial performance advantages over its predecessor. Announced in March 2025, the chip offers 50% more memory and FP4 compute capability than the B200. Nvidia marketed the B300 as built for the age of reasoning, specifically referencing sophisticated large language models like DeepSeek R1. The chip was originally expected to ship by the end of 2025, and it quickly became a highly sought-after asset in the AI market.

The smuggling of these high-value components has led to legal action in recent months. In August 2026, nine individuals in Taiwan were indicted for the illegal export of B300 GPUs to China. This indictment came shortly after Supermicro fired several employees following an internal investigation into smuggling activities. The complex operation involved a Taiwanese server trading firm called Flying Tiget Tech, which held whitelisted status for purchasing GPUs. This case illustrates the lengths to which intermediaries will go to move controlled hardware across borders.

The scarcity of Nvidia's latest chips has created a ripple effect across the hardware market. Demand for the company's older silicon has surged, driving prices to record levels. For instance, A100 servers, which are five years old, have tripled in price and can now cost up to $82,000. Some organizations have even resorted to retooling gaming GPUs to handle AI workloads when data center chips are unavailable. This dynamic underscores the critical importance of Nvidia's products in the current AI infrastructure landscape.

Nvidia is currently working on a new tool that would allow its chips to be located more easily, though it has not yet been released. Until then, vulnerabilities in existing monitoring software like the DCGM Exporter may continue to provide insights into the location of restricted hardware. The discovery by Lava serves as a reminder that while export controls are a primary mechanism for managing technology flow, they are not an impenetrable barrier against determined actors in the global market.