Los AngelesNew YorkChicagoHoustonPhoenixPhiladelphiaSan FranciscoSeattleAtlantaBoston
Saturday, October 24, 2026The stories that matter, every day.
Business · AI

B300s Found in State Telecom Network

Israeli security firm detects 32 banned Nvidia chips in China Telecom’s infrastructure, prompting the GPU maker to downplay the breach as a minor issue.

By Christopher Reed ·

Thirty-two Nvidia Blackwell Ultra B300 graphics processing units have been identified within the network of China Telecom, a state-owned telecommunications provider. The discovery, made by Israeli security firm Lava, highlights a persistent gap in export controls. These specific chips are barred from export by U.S. regulations, meaning they should only be sold to whitelisted buyers who agree not to pass them to sanctioned parties, including entities in China.

The timeline of this leakage reflects a broader struggle for supply chain security. In 2025, reports indicated that $1 billion worth of Nvidia AI chips were smuggled into China in just three months. While the B200 was the primary target at that time, distributors were already positioning themselves to move the newer B300 series once it became available. The B300 was announced in March of last year, touted as a chip built for the age of reasoning with 50% more memory and FP4 compute than its predecessor.

The detection of these chips raises questions about the vulnerability of monitoring tools. Lava identified the B300s using Nvidia’s own DCGM Exporter tool, a software designed for internal health and performance monitoring by data center operators. The tool can often be accessed beyond the immediate data center network and does not always require password authentication. It is currently unknown exactly where within China Telecom’s infrastructure the chips are located or how they arrived there.

Nvidia has responded to the findings by characterizing the incident as statistically insignificant. A spokesperson stated that governments and researchers have found diversion of controlled products in recent court cases to be less than half of one percent. The company described this as a drop in the bucket compared to the vast amount of domestic compute China already possesses. With tens of millions of GPUs in service in approved countries, Nvidia maintains that smuggling remains a non-factor, blocked by governments on both sides of the globe.

The stakes for these chips are high due to their performance capabilities. The B300 is one of the most powerful data center GPUs available, and its availability has made it a hot property. In August 2026, nine people in Taiwan were indicted over the illegal export of B300 GPUs to China. This followed days after Supermicro confirmed it had fired several employees following an internal investigation into smuggling. The illicit trade involved a Taiwanese server trading firm called Flying Tiget Tech, which held whitelisted status for GPU purchases.

The scarcity of high-end silicon has driven up prices for older inventory. Nvidia’s A100 servers, which are five years old, have seen their prices triple, reaching up to $82,000. The intense demand has led some operators to retool gaming GPUs to handle AI workloads where possible. This economic pressure underscores why Chinese firms are so keen to acquire the latest Blackwell Ultra hardware despite the regulatory hurdles.

Nvidia is working on a new tool that would allow its chips to be more easily located, though it has not yet been released. Until then, the reliance on existing monitoring software like the DCGM Exporter leaves potential gaps in visibility. The company did not deny Lava’s findings but insists the volume of smuggled chips does not significantly impact the overall market dynamics.

The next phase involves monitoring how these chips are utilized within China Telecom’s operations. While Nvidia downplays the financial impact, the presence of 32 B300s in a state-owned network serves as a tangible example of the challenges in enforcing export controls on critical AI infrastructure.

Source: Tom's Hardware

More from AI