The United States sharply expanded export controls on advanced semiconductors and chipmaking equipment this week, closing pathways that allowed technology companies to design China-specific products around restrictions imposed a year ago. The Commerce Department’s Bureau of Industry and Security said Tuesday that the revised rules will capture additional artificial-intelligence chips, broaden controls on semiconductor manufacturing equipment and make it harder for Chinese companies to obtain restricted processors through overseas affiliates.

The new framework directly affects Nvidia products developed after the October 2022 controls. In a Securities and Exchange Commission filing, Nvidia said the licensing requirements cover products including the A100, A800, H100, H800, L40, L40S and RTX 4090, as well as systems incorporating covered chips. The filing shows how rapidly export-control thresholds can alter product strategy in the AI market: the A800 and H800 had been created in part to comply with the previous restrictions.

Commerce targets performance and circumvention

BIS said the changes are intended to preserve the national-security effect of the 2022 rules as chipmakers change architectures and customers seek alternative supply routes. The agency’s October 17 announcement replaces one of the earlier technical thresholds with a performance-density test, adds notification requirements for some chips just below the main control line and expands licensing requirements to companies headquartered in arms-embargoed countries or whose ultimate parents are based there.

The underlying advanced-computing interim rule lays out the technical revisions to the Export Administration Regulations. It preserves restrictions on high-performance integrated circuits while revising the criteria used to determine which processors require a license. BIS says the goal is to reduce opportunities to engineer narrowly compliant chips that retain enough computing power to support advanced military or intelligence applications.

The policy remains focused on high-end computing rather than ordinary consumer semiconductors. Commerce Secretary Gina Raimondo said most chips will continue to trade without restriction, while the government concentrates controls on processors and manufacturing capabilities with the greatest potential military relevance.

Chipmaking equipment faces broader controls

The package is not limited to Nvidia-style accelerators. A separate semiconductor-equipment rule expands controls on tools used to fabricate advanced integrated circuits and refines restrictions on U.S. persons who support certain advanced Chinese semiconductor facilities. It also extends some licensing requirements beyond China and Macau to other countries subject to U.S. arms embargoes.

That manufacturing dimension is important because restricting finished AI chips does not by itself prevent China from developing domestic alternatives. The U.S. strategy therefore targets two points in the supply chain: access to the most capable foreign processors and access to equipment needed to manufacture advanced chips independently.

BIS also placed additional Chinese entities involved in advanced computing on the Entity List. The agency’s Entity List rule adds two Chinese companies and subsidiaries—13 entities in total—and applies foreign direct product restrictions intended to prevent them from obtaining chips made abroad with U.S. technology without a license.

Nvidia becomes an immediate test case

The commercial impact is clearest at Nvidia, whose graphics processors have become core infrastructure for training and running generative-AI systems. The company’s A100 and H100 are among the industry’s most sought-after accelerators, while the A800 and H800 were lower-bandwidth variants built for the Chinese market after last year’s controls.

The revised rules make clear that simply reducing interconnect bandwidth is no longer sufficient. Nvidia told investors that the new requirements apply to a broader set of products and destinations and can extend to customers headquartered in countries subject to U.S. arms embargoes even when the physical shipment goes elsewhere.

The Associated Press reported that Commerce is attempting to close loopholes while preserving trade in less sensitive chips. Raimondo described the effort as a national-security measure rather than an attempt to halt all semiconductor commerce with China. That distinction is central to the administration’s policy: a narrow set of technologies is being treated as strategically consequential enough to justify licensing restrictions even at significant cost to U.S. suppliers.

China objects as the technology boundary moves again

Beijing responded Wednesday by accusing Washington of abusing export controls and disrupting global supply chains. The Associated Press reported that China’s Commerce Ministry said the new measures cause losses to semiconductor companies and interfere with normal trade, while promising unspecified steps to protect Chinese interests.

The dispute illustrates why export controls are becoming an iterative policy instrument. Computing performance is advancing too quickly for static thresholds to remain effective indefinitely. Companies can redesign products, buyers can route transactions through affiliates and domestic Chinese manufacturers can improve. BIS is therefore signaling that it intends to revise the rules as technology and circumvention methods evolve.

For the technology industry, that creates a new kind of regulatory risk. Product road maps for AI accelerators can now be altered not only by engineering constraints and market demand but also by national-security definitions of which combinations of processing power, density and end use are acceptable for export.

As of Saturday, the practical effect is clear: Washington has widened the fence around the most advanced AI computing technology available to China. Whether the controls meaningfully slow Chinese military and AI capabilities—or accelerate domestic substitution—will depend on enforcement, allied cooperation and how quickly semiconductor technology moves beyond the thresholds written this week.