Nvidia: 'We don't install secret tracking devices in our products' — GPU giant responds after Washington accused of secretly tracking AI server shipments at risk of diversion to China
Nvidia wants no part in the story.

Update Aug 14 6:30 am ET: The opening of this story has been updated following a statement from Nvidia in response to accusations of chip tracking levelled at the Federal government.
Nvidia has issued a flat denial of the existence of tracking chips in its hardware, following reports that the Federal Government allegedly placed tracking devices in shipments of AI server tech thought to be at risk of diversion to China. Earlier this week, Reuters reported that anonymous sources claim U.S. authorities are employing this tactic as part of their investigation into the alleged smuggling of banned AI chips into China. In a statement to Tom's Hardware, an Nvidia spokesperson simply stated, "We don't install secret tracking devices in our products." While the story in question lays the tracking at the feet of Washington, rather than any companies whose products are involved, Nvidia's staunch denail reflects a desire to distance itself from reports as it prepares to resume selling H20 chips in China.
According to the report, the alleged tracking doesn't impact all shipments, but only "targeted" shipments believed to be at a high risk of being diverted to China. Despite the affordability and ubiquity of trackers, it will be prohibitively expensive to place several in all servers that are going through customs. It’s more likely that the authorities are allegedly doing this to aid in a specific investigation. Although placing a tracker on a package typically requires a court order, Reuters notes that export enforcement agents may sometimes do so with administrative approval only.
It’s unclear when this alleged practice started for semiconductors, but those involved in the AI chip supply chain seem aware of these actions. These trackers are reportedly often added to the shipping containers, the packaging of individual servers, or even within the server racks themselves. The report cites individuals who claim that they had seen images and videos of trackers being removed from Dell and Super Micro servers, some of them as large as smartphones.
Smugglers also apparently know this, with court documents in a recent case of smuggling millions of dollars' worth of chips to China revealing that a suspect instructed their co-conspirators to “pay attention to see if there is a tracker on it.”
The U.S. Department of Commerce, Homeland Security, and the FBI didn’t issue a statement when asked by Reuters, although the sources confirm that these agencies are typically involved in these situations.
Nvidia CEO Jensen Huang has consistently denied that the company’s most advanced AI chips are being diverted to China. However, multiple outlets confirm that there’s a strong black market for these cutting-edge silicon, with a reported billion dollars’ worth of Nvidia products being openly traded into the East Asian country in just the last quarter.
Washington is trying to hamper China’s progress in AI, especially as it’s considered a ‘dual-use technology’ that is applicable in both civilian and military purposes. Because of this, the White House — across multiple administrations — has applied export controls on state-of-the-art semiconductors, including the most powerful AI chips from Nvidia and AMD. However, Huang has said that the U.S. export control was a failure, with the former U.S. Commerce Secretary also saying that holding back China’s chipmaking progress is a fool’s errand.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Despite that, the Trump administration has doubled down on its efforts to control the flow of AI chips into China, even going as far as banning both the Nvidia H20 and the AMD MI308 in April 2025 — AI chips that have been especially designed to comply with the initial set of export controls released by the Federal government.
The U.S. has since backed down on this particular export control and has reopened the taps for these AI chips; the newer, more powerful variants of these semiconductors still aren’t allowed to be shipped to China. In the most recent twist in the saga, AMD and Nvidia have struck a deal to share 15% of related sales revenue with the government in exchange for export licenses for its H20 and MI308, an unprecedented arrangement.
The U.S. has been clamping down on the alleged smuggling of high-end semiconductors. American allies such as Singapore are taking action to stop the flow of these AI chips through their borders. Furthermore, a bill has been filed in Congress to force American AI chip makers to add geo-tracking technology into their products. Nvidia is against this, though, saying that these ‘permanent flaws’ would make their devices vulnerable to hackers. These developments have nonetheless led the Chinese government to question the integrity of the company’s chips and have even discouraged their use.
Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Jowi Morales is a tech enthusiast with years of experience working in the industry. He’s been writing with several tech publications since 2021, where he’s been interested in tech hardware and consumer electronics.
-
TechieTwo IMO there is nothing wrong with tracking controlled chips, especially with so much fraud these days.Reply -
jp7189 I saw that gamers nexus is using their contacts to do a deep dive on the smuggling of gpus. I believe they are supposed to release their finding on the 15th. Not a usual source and not their usual bread and butter, but should be interesting to hear.Reply -
zsydeepsky Well, just imagine if Huawei (or any other Chinese company) did anything remotely like this.Reply
The report and the comments will be in a very different tone, for sure. -
bolweval
How do you know they haven't?zsydeepsky said:Well, just imagine if Huawei (or any other Chinese company) did anything remotely like this.
The report and the comments will be in a very different tone, for sure. -
bit_user I think this is too easily circumvented.Reply
IMO, the best way is for the GPU to have an embedded license key and a built-in counter. The counter tracks the number of hours the GPU has been in use. Once it reaches the limit, for the license key, the GPU goes into "limp mode" and won't run above idle clocks or something. You have to load in a new license key. Then, their network of license key servers can measure ping time between the GPU and the nearest couple license servers, in order to ensure the GPU is located within range.
Assuming the counter is embedded sufficiently deep inside the GPU, and you have some redundancy or tamper-detection, this could be highly-effective because it would basically enforce range limits on the GPU's location, over the span of its lifetime. From the customer's perspective, it wouldn't require continual connectivity to the license server, only periodic connections. That should be more palatable to them, from a network security point of view, because you could just reconfigure the networking of a machine, only when it needs to be relicensed, and then put it back on a private segment to continue its work. -
edzieba I would doubt anyone is installing trackers in servers.Reply
However, installing trackers in containers, or in individual pallets, for shipment? That isn't even uncommon for high-value shipments, and these DGX racks can be 6-8 figures each depending on configuration. Yes, sellers (Nvidia or otherwise) absolutely want to know where those are between dispatch and receipt. -
-Fran- The wording is funny. I'm sure they don't go out of their way to "install" anything, but just can track the hardware organically.Reply
If it's connected to the internet, it can always "ping home" in some capacity since, at the end of the day, you still need their drivers to make it work and very few have the capacity to write full fledged drivers for these GPUs or even try to modify them. This being said, we already saw this with the LHR cards in the height of the mining boom.
Super cynical take, but I just can't give nVidia the benefit of the doubt.
Regards. -
TJ Hooker
Why not read the article rather than speculate? It is quite clearly talking about adding separate tracking devices, sometimes as large as cell phones, to packaging. Not tracking functionality being built into GPUs themselves. The latter would be trivially easy to detect if it was just pinging home over the internet, so if no one has seen that it's probably not happening (at least not yet).-Fran- said:The wording is funny. I'm sure they don't go out of their way to "install" anything, but just can track the hardware organically.
If it's connected to the internet, it can always "ping home" in some capacity since, at the end of the day, you still need their drivers to make it work and very few have the capacity to write full fledged drivers for these GPUs or even try to modify them. This being said, we already saw this with the LHR cards in the height of the mining boom.
Super cynical take, but I just can't give nVidia the benefit of the doubt.
Regards. -
-Fran-
I know they were talking about "proper" trackers. I just went on the tinfoil tangent on purpose.TJ Hooker said:Why not read the article rather than speculate? It is quite clearly talking about adding separate tracking devices, sometimes as large as cell phones, to packaging. Not tracking functionality being built into GPUs themselves. The latter would be trivially easy to detect if it was just pinging home over the internet, so if no one has seen that it's probably not happening (at least not yet).
There's many different ways a device can "ping" back home which do not include a direct "call" to a location via the same hardware, just sayin'. Plenty creative ways you can accomplish it.
Regards.