Can Location Verification Stop AI Chip Smuggling?
A New Bill Aims to Track Where AI Chips End Up, But Will It Be Enough?
The U.S. has spent the past few years tightening the leash on its most powerful tech export: advanced AI chips. These aren’t your typical GPUs powering gaming rigs. We’re talking about Nvidia’s H100s—the crown jewels of modern machine learning. The logic is simple: limit China and other adversaries' access to this silicon, and you slow down their ability to develop cutting-edge AI.
But there's a problem. Despite the restrictions, tens of thousands of these chips are slipping through the cracks every year. Enter the Chip Security Act.
On May 9, Senator Tom Cotton (R-AK) introduced legislation that would require location verification for export-controlled chips. A bipartisan companion bill followed in the House days later. The core idea: embed a kind of digital homing beacon in every AI chip that leaves the U.S.
This push comes amid efforts by the Trump administration to replace Biden-era chip export rules with a framework that is, in the words of OSTP director Michael Kratsios, “strict and simple.” AI and crypto czar David Sacks put it more bluntly: trust, but verify.
The Problem: Shadow Markets and Shell Companies
Despite 2022's export controls, AI chips are still getting to China. A new report by the Center for a New American Security (CNAS) estimates that up to 100,000 restricted GPUs may have entered China illegally last year. Some think it could be as high as one million.
How? Shell companies in non-restricted countries. Shady middlemen. Weak enforcement.
Bloomberg uncovered a $300 million smuggling operation via an Indian pharmaceutical company. The New York Times found brokers moving $100 million worth of chips through Shenzhen. Even DeepSeek, one of China’s AI hopefuls, claims to have 10,000 Nvidia chips; some say it could be closer to 50,000.
A Solution: Delay-Based Location Verification
Location verification, as proposed in the Chip Security Act, could close a critical blind spot. But how do you actually verify where a chip is without compromising privacy or relying on GPS, which can be spoofed or blocked by concrete walls?
The answer might be something called delay-based location verification. It works like this:
A global network of “landmark” servers pings AI chips.
Chips respond.
The round-trip delay is measured to estimate the chip’s distance from the server.
Triangulate responses from multiple landmarks and you get a rough-but-useful estimate of location—enough to know whether a chip is still where it’s supposed to be.
Former Google CEO Eric Schmidt backed a similar approach earlier this year in a policy paper.
Privacy-Friendly, Sabotage-Resistant?
Critics are quick to raise the specter of the Clipper Chip—a 1990s-era proposal for government-accessible encryption backdoors. But supporters argue this is different. Delay-based location checks don’t expose data or customer behavior. They don’t even reveal the exact chip location. Just whether it’s in, say, Singapore or somewhere in central China.
What about tampering? Could hackers disable this beacon? It won’t be easy. According to experts, Nvidia’s secret key is etched onto the chip die itself. Removing or altering it would require cutting-edge equipment and serious risk.
And even if someone managed to spoof the timing, the inconsistencies in response latency would likely serve as a red flag. Imagine a chip claiming to be in Switzerland but sending signals that imply it's somewhere between Alaska and New Zealand.
Is It Feasible?
The good news: implementing this might not require a hardware overhaul. Nvidia already has the comms hardware built-in. Firmware updates could activate location verification capabilities.
Global coverage could be achieved with fewer than 500 landmark servers. Targeted deployments—say, near known smuggling hotspots—would need even less. Cloud providers could host the infrastructure, meaning the upfront cost would be manageable.
Nvidia itself has said transmitting limited telemetry is technically viable.
Not a Silver Bullet, But a Strong Start
Location verification won’t magically end chip smuggling. The Bureau of Industry and Security (BIS), the agency enforcing these controls, is underfunded and understaffed. Smugglers are resourceful, and geopolitics is messy.
Still, this is a powerful tool in the toolbox. It increases the friction, raises the stakes, and makes enforcement less like whack-a-mole. In combination with other tools like geofencing (disabling chips in unauthorized regions), it could help stabilize a system where too many high-value assets are vanishing into the black market.
More importantly, it could give U.S. regulators the visibility they need to make smarter decisions about where, how, and to whom these chips are sold. And that’s a move toward not just controlling AI, but understanding the battlefield on which the next tech cold war will be fought.
For more on the AI chip race and evolving export controls, subscribe to AI Innovations Lab’s AI-READY Newsletter.