Nvidia’s China Partners and the PLA

Nvidia doesn't sell directly to the Chinese military. It doesn't have to. According to research shared by CSET's Sam Bresnick in The Wire China, a web of resellers, cloud providers, and systems integrators in China has moved Nvidia's AI chips into the hands of organizations tied to the People's Liberation Army and other entities the U.S. has explicitly restricted.

This isn't really a story about Nvidia breaking rules. It's a story about how porous those rules are once hardware crosses into a market with thousands of intermediaries, shell-like distributors, and cloud resale arrangements that make end-use nearly impossible to police from Santa Clara. A chip sold legally to a data center operator in Shenzhen can quietly end up powering a defense research lab three transactions later, and nobody along that chain necessarily has to lie outright to make it happen.

Bresnick's point, stated plainly, is that current U.S. export controls leave real gaps. Washington has spent years tightening restrictions on advanced GPUs bound for China, adjusting thresholds, closing loopholes on chip performance, and pressuring allies to fall in line. Yet the enforcement problem has always been less about the rules on paper and more about tracking hardware after it leaves a warehouse. China's commercial AI ecosystem is large, fast-moving, and deeply intertwined with state and military-linked institutions in ways that don't map cleanly onto a compliance checklist.

And that's the uncomfortable part for policymakers. Tightening chip export rules further sounds decisive, but it does little if the actual leakage happens through obscure resale networks rather than direct Nvidia-to-military sales. The fix, if there is one, probably involves far more aggressive tracking of downstream buyers and partners, not just another round of restrictions on the chips themselves.