Sales of Nvidia’s H200 AI chip to China approved under new US security conditions

A conditional greenlight for powerful AI hardware

The Trump administration has effectively allowed Nvidia to sell its H200 artificial intelligence chips to Chinese buyers, while simultaneously adding new security requirements intended to reduce national security risks. The Commerce Department’s Bureau of Industry and Security set conditions that include third-party review of chips before export and requirements aimed at ensuring the U.S. market is adequately supplied.

The decision underscores an ongoing tension: U.S. officials want to protect sensitive technology that could enhance an adversary’s military or cyber capabilities, but they also face pressure to let American firms compete globally and maintain scale in a market where demand for AI compute is accelerating.

What the new requirements try to do

Under the new approach, China is barred from using the chips for military purposes, and the rules place limits tied to the relationship between chips sold to U.S. customers and exports to China. The framework aims to prevent a scenario where large portions of U.S.-oriented supply effectively become an export pipeline, while also establishing checks meant to confirm the products and customers comply with the licensing terms.

Nvidia argued that allowing sales to vetted commercial customers, overseen through the licensing process, strikes a balance between security and competitiveness. Supporters of the decision say it helps preserve U.S. jobs and investment at a time when the country is trying to strengthen domestic semiconductor manufacturing and related supply chains.

Why critics remain concerned

Some Democratic lawmakers have warned that even chips that are not the company’s absolute top tier can still meaningfully boost advanced computing capacity. They argue that AI accelerators can be used for military planning, signals processing, surveillance, and offensive cyber operations, and that enforcement can be difficult once chips enter complex global supply networks.

The H200 is not Nvidia’s newest platform—those include newer architectures and future roadmaps—but it remains a high-performance accelerator. That means the practical question for policymakers is not just “Is it the best chip?” but “Is it enough to change capability at scale?”

A broader export-control strategy in motion

The approval fits into a broader pattern of U.S. attempts to calibrate export controls: limiting the most advanced items while allowing some commercial sales under licensing and oversight. The policy debate is likely to continue as AI models and hardware evolve quickly, and as competitors seek alternative supply routes or develop domestic substitutes. For now, the decision signals that Washington is trying to manage, rather than fully sever, parts of the AI semiconductor trade with China.