By Hamza Faisal

Nvidia is tightening its grip on where its most powerful AI chips end up. The chipmaker now requires cloud providers in Southeast Asia that want its partner status to enrol in a tracking system that can pinpoint the location of its accelerators, according to a Bloomberg report published on November 6.

The requirement applies to companies seeking Nvidia Cloud Partner status in the region, a designation that brings public recognition and preferred access to Nvidia's hardware. To qualify, applicants must sign up for Fleet Intelligence, Nvidia's monitoring service, which collects telemetry and geolocation data on individual components.

The move comes as US authorities investigate claims that restricted AI processors are reaching China through overseas intermediaries, with Southeast Asia seen as a key transshipment route. Officials believe smuggling networks have moved billions of dollars of hardware into China despite export controls that broadly bar Chinese companies from buying advanced Nvidia chips without a licence.

A voluntary tool becomes a condition

Fleet Intelligence began life nearly a year ago as a data-centre management tool. It uses a software agent installed on GPU systems to stream operational data — utilisation, hardware health and configuration — to Nvidia's cloud service, and it can estimate a processor's location using telemetry and network-latency measurements.

Until now, participation has been voluntary. The new rule turns enrolment into a condition of doing business as an official Nvidia partner in Southeast Asia, giving the company a commercial lever to widen adoption. Nvidia has not said whether existing cloud partners in the region will have to enrol to keep their status, or whether the requirement will spread to other markets. It declined to answer Bloomberg's questions on both points.

There are limits to what the system does. Enrolling does not automatically require customers to disclose GPU locations to Nvidia or to US regulators, and the company says it does not receive location data by default. What it gains is visibility: a way to investigate whether chips have been diverted after they leave the supply chain.

The smuggling route behind the rule

The scrutiny has intensified around Megaspeed International, a company under investigation in the United States and Singapore over its data-centre footprint. Nvidia chief executive Jensen Huang has said the company studied Megaspeed's deployments and counted every processor, though Nvidia declined to say whether it carried out further checks after earlier concerns.

US authorities have reportedly also placed location-tracking devices in select shipments of AI servers to investigate possible diversions to China.

The squeeze tightens on both sides

The partner-status rule is not Nvidia's only new control. The company has reportedly cut about half of the Asian companies it had previously pre-approved for chip purchases, and it has told cloud buyers not to rent more than half of their Nvidia computing capacity to Chinese customers — a 51 per cent rule that limits how overseas operators can serve Chinese clients even when the hardware sits outside China.

At the same time, supply is tightening from the other direction. Around 600,000 H200 chips are said to be stockpiled in Hong Kong, awaiting shipment to mainland China while Beijing withholds approval for large-scale imports — a reminder that the chip war now runs on two fronts: Washington restricting what leaves, and Beijing restricting what comes in.

For Nvidia, the balancing act is delicate. It must show regulators it is policing diversion without alienating the cloud customers that drive its data-centre business. Fleet Intelligence, built as a management tool, has now become its answer — a tracker that is technically voluntary, commercially hard to refuse, and politically impossible to ignore.