U.S. Action: Plans to Block Chinese Companies from Remotely Accessing Computing Power Through Overseas Data Centers
On November 17, news reported that the U.S. government is further tightening export controls on advanced artificial intelligence chips, attempting to close previous regulatory loopholes and prevent companies in restricted countries such as China from circumventing direct export bans by remotely accessing the computing power of U.S. advanced AI chips deployed in data centers in third countries.
On November 16 local time, the U.S. Department of Commerce officially launched a temporary regulatory rule, clearly requiring all relevant companies to apply for a special license with regulatory authorities in advance whenever exporting, re-exporting, or transferring AI chips and supporting servers to data centers worldwide, as long as the relevant facilities have the possibility of providing computing power services to enterprises in China or other restricted countries.
This means that the scope of this control has completely gone beyond previous boundaries. It is no longer limited to whether chips and servers are physically imported into restricted countries; even the entire chain of actions, where chips are fully deployed in overseas third countries and subsequently provide computing power to restricted clients through cloud services, is also included within the strict regulatory scope.
These new rules target the D:5 country group explicitly defined in the U.S. export control system, which includes countries such as China, Russia, Iran, and North Korea, all of which are subject to U.S. arms embargo restrictions. Prior to this, the U.S. had already explicitly required all companies to obtain export licenses in advance when selling advanced AI chips to entities in these countries or to companies actually controlled by these countries' enterprises.
The adjustment rules introduced this time further extend the regulatory reach to all scenarios in which overseas data centers provide remote computing power services. In the interim rules, the U.S. Department of Commerce set the threshold for computing power applicability at more than 12 million total processing performance (TPP).
According to estimates by the overseas technology media TechCrunch, this computing power scale is roughly equivalent to the maximum computing capability output by a total of 750 NVIDIA H100 or H200 chips, which is approximately enough to support the full-chain training needs of a medium-sized large model.
The regulatory boundaries of the new rules are quite broad. Even if Chinese companies do not directly sign contracts with overseas data centers, but instead obtain the corresponding computing power indirectly through intermediaries, agents, or other multi-layered third-party entities, they will still fall under the regulatory oversight of the new rules.

In order to implement this set of regulatory requirements, all chip exporters and data center operators worldwide must significantly upgrade their customer due diligence processes, verifying the ownership structure, actual control relationships, and all business activity flows of each customer one by one, to strictly prevent companies from restricted countries from bypassing control restrictions through complex multi-layered corporate structures or intermediary agencies.
If an overseas data center was already providing computing power services that meet the threshold to enterprises in relevant countries before the new regulations officially take effect, it must also apply for the corresponding special permit within 90 days after the regulations take effect. Entities that fail to complete the application within the time limit will subsequently be directly included in the U.S. export restriction list.
The new regulations additionally include Malaysia, Singapore, Indonesia, Thailand, and Brazil within the scope of advanced semiconductor export licensing requirements.
The reason given by the U.S. Department of Commerce is that the risk of abnormal transit of advanced chips in these countries is far higher than the global average.
However, the rules also include exemptions. If the chip is ultimately delivered to a company headquartered in the United States, or to a company whose ultimate parent company's headquarters is registered in the United States, it may not be subject to this new licensing requirement, provided that the conditions for periodic reporting required by regulations are met.
These new regulations directly extend the United States' AI chip control boundaries from the physical flow of chips to the flow of computing power services, effectively using long-arm jurisdiction to bring almost all high-level AI computing resources worldwide under U.S. regulatory oversight, nearly blocking all channels of overseas high-end computing power accessible to Chinese companies.
But this kind of extreme regulatory action that breaks the boundaries of marketization will, in turn, only force the accelerated implementation of a fully domestic AI computing power system, and ultimately will entirely surrender the global market space of high-end American chips.
