
The United States has moved to close a gap in its advanced chip controls by targeting remote access to American computing power housed in overseas data centres. The Commerce Department’s Bureau of Industry and Security (BIS) announced an interim final rule on 16 November aimed at arrangements that could let restricted customers use cutting-edge processors without importing the hardware themselves.
The distinction matters. Existing export controls largely follow the physical movement of chips. A company that cannot obtain them directly may still seek computing capacity from a cloud provider in another country. BIS says the new rule addresses that route by imposing requirements on certain overseas data centres and the entities providing access to advanced computing resources.
According to the BIS announcement and its published rule text, the measure applies to specified advanced computing items at data centres outside the United States when they serve firms headquartered or ultimately parented in countries covered by the agency’s D:5 group or Macau and meet the rule’s thresholds. Operators would face due-diligence obligations and, in relevant cases, licensing requirements. The rule is not a blanket prohibition on all overseas AI computing or ordinary cloud use.
BIS also sets out additional controls for certain advanced semiconductors destined for Indonesia, Malaysia, Singapore, Thailand and Brazil, while providing conditional treatment for US-headquartered operators. The details and exceptions will matter to cloud companies, chip suppliers and data-centre developers planning capacity in those markets. Firms will need to examine the final published text and effective date before treating the announced requirements as already in force.
The rule follows concern that access to high-end AI infrastructure is increasingly sold as a service, rather than as a shipment of hardware. TechBooky previously reported on Chinese AI firms seeking to rent Nvidia computing power abroad. This announcement is a policy response to that wider pattern, though it does not establish that every such rental breaches US law.
For providers, the difficult question is not simply where a server sits. It is who controls the customer, which chips are being used, what level of capacity is supplied and what checks are needed before access is granted. That could raise compliance costs and slow some cross-border contracts. It may also influence where new AI data centres are financed, especially in regions that have attracted investment partly because they can serve clients across multiple jurisdictions.
The broader contest is over whether Washington can keep its chip restrictions effective as AI computing becomes more distributed and easier to rent. The new BIS rule marks a significant attempt to extend oversight from hardware exports to access to the computing power those chips deliver. Its practical reach will depend on implementation, licences and how providers verify their customers.







