Nvidia mandates monitoring for Southeast Asian cloud partners to curb China smuggling

2–3 minutes

Nvidia has begun requiring new cloud partners in Southeast Asia to adopt its Fleet Intelligence monitoring system as a prerequisite for obtaining the Nvidia Cloud Partner certification. This move aims to prevent the illegal diversion of advanced chips to China, according to a Bloomberg report.

How Fleet Intelligence works

The software, launched in May, is installed on servers and transmits real-time data to a cloud-based dashboard. Operators can monitor metrics such as usage, temperature, power consumption, and failures, and also view the physical layout of the equipment. The tool does not interfere with GPU functionality and is provided free of charge to both equipment owners and customers who rent processing capacity.

The context of US investigations

A complaint filed by the US Department of Justice in March outlined a scheme in which servers purchased by an intermediary company were transported from the US to Taiwan and Southeast Asia, and then shipped to China in unmarked boxes. Between 2024 and 2025, the intermediary acquired approximately $2.5 billion worth of equipment, of which around $510 million was diverted to China within a few weeks in late April to mid-May 2025. The allegations are currently under judicial review.

Implications for the AI infrastructure market

The monitoring requirement adds to existing restrictions imposed by export licenses and supply chain challenges. Recent reports suggest that Chinese buyers expect to receive only a small fraction of the chips they ordered. Beyond controlling who acquires the equipment, the new policy raises questions about the location of chips after sale, particularly in regions hosting data centers that provide remote services to global clients.

What it means to be an Nvidia Cloud Partner

Nvidia’s partner program comprises providers offering cloud computing capacity based on the company’s technology. Customers can hire these services to train or run artificial intelligence models without needing to purchase and maintain their own servers. Companies like Lambda and Nebius are already part of the network, selling access to infrastructure installed in their data centers.

The adoption of Fleet Intelligence is not retroactive; existing certified partners are exempt from the new rule, and the program remains voluntary outside of the certification requirement for new Southeast Asian members.