U.S. AI Chip Rules Target a New China Access Loophole

The United States is moving to tighten its controls on advanced AI chips, targeting a potential loophole that may have allowed Chinese companies to access cutting-edge processors through subsidiaries outside China.

The new guidance from the U.S. Department of Commerce is aimed at advanced chips such as Nvidia’s Rubin and Blackwell processors and AMD’s MI350x, Reuters reported. The concern is that Chinese-headquartered companies could obtain powerful AI chips through operations in other countries, including regions such as Malaysia.

The move highlights how difficult AI chip controls have become. The most advanced processors are no longer just hardware components. They are strategic infrastructure for training and running large AI models, powering data centers and supporting national technology ambitions.

For Washington, the issue is not only where a chip is shipped. It is who ultimately controls the company buying it.

According to Reuters, the new approach requires licenses for advanced chip sales to Chinese-headquartered entities regardless of where those entities are physically located. That means a Chinese company’s overseas subsidiary could face the same restrictions as an entity operating inside China.

That distinction matters because modern AI supply chains are global. A technology company may have headquarters in one country, subsidiaries in another and data center operations in a third. If export rules focus only on geography, companies may be able to structure purchases through locations that are not directly covered by restrictions.

The latest guidance appears designed to close that gap.

The U.S. has been trying for years to limit China’s access to the most advanced AI and semiconductor technology. The goal is to slow China’s ability to build high-end AI systems that could support military, surveillance or strategic computing programs. China, meanwhile, has been pushing to build more domestic chip capacity and reduce dependence on foreign suppliers.

That competition has made companies such as Nvidia and AMD central to a wider geopolitical struggle. Their most powerful AI chips are critical for data centers and large-scale model development, but they are also subject to increasingly complex export rules.

The new guidance may affect more than chipmakers. Cloud providers, data center operators, server manufacturers and AI startups could all need to pay closer attention to ownership structures, customer identity and licensing requirements.

The Reuters report also noted that the guidance does not require data centers to stop using or servicing affected chips already in place. That detail is important because the rule appears focused on future shipments and access pathways, not an immediate shutdown of existing infrastructure.

For AI companies, the practical effect could be more uncertainty. A company that once planned to buy chips through an overseas subsidiary may now need a license or may have to redesign its infrastructure plans. That can slow deployment, raise costs and push firms toward alternative chips.

For Chinese companies, the pressure may accelerate the search for domestic AI processors. China has already been promoting local alternatives, though advanced chip manufacturing remains a challenge. Some domestic chips may be good enough for certain AI workloads, but matching the highest-end Nvidia and AMD systems remains difficult.

For U.S. companies, the issue is more complicated. Export controls can protect strategic technology, but they can also limit sales in one of the world’s largest markets. Nvidia and AMD have already had to adapt products and sales strategies around U.S. rules.

The latest move also shows how fast AI policy is evolving. Rules written one year can create unexpected workarounds the next. As companies change where they operate and how they source chips, governments have to update enforcement strategies.

This creates a constant cycle: restrictions, adaptation, new guidance and further restrictions.

For ordinary users, the story may seem distant. But AI chip access affects the tools people eventually use. The availability of advanced processors shapes how quickly AI systems can be trained, how cheaply they can run and which companies can compete at the highest level.

The U.S. guidance is therefore not just about one shipment path. It is about who gets to build the next generation of AI infrastructure.

The broader message is clear: advanced AI chips are becoming a controlled resource. As governments treat compute power as a strategic asset, the global AI race is moving deeper into trade policy, supply chains and data center regulation.

The new rules may not end the struggle over AI chip access. But they show that Washington is now looking beyond borders on a map and focusing more closely on corporate control, ownership and the real end user behind each purchase.