The US Commerce Department is extending export controls to remote access to advanced AI semiconductors, closing a channel that allowed Chinese companies to rent computing capacity at data centers outside China. The Bureau of Industry and Security issued an interim final rule on Monday, with an effective date of approximately November 16, 2026. Data centers providing significant compute to Chinese entities, including activity involving China, Macau or certain other restricted jurisdictions, must obtain a license and conduct enhanced customer due diligence after exceeding 12 million aggregate total processing performance, equivalent to approximately 760 Nvidia H100 GPUs. Operators have 90 days after the rule takes effect to comply and may face future restrictions if they do not. Exports of advanced AI chips to Indonesia, Malaysia, Singapore, Thailand and Brazil will also require licenses. US-headquartered operators are exempt from certain licensing requirements but must file monthly compliance reports. The measure follows allegations surrounding Nvidia and data-center operator Megaspeed, which bought at least $4.6 billion of Nvidia servers between 2023 and 2025 and was accused of enabling diverted hardware and remote access for major Chinese clients including Alibaba. Shares of Southeast Asian companies linked to data centers fell after the announcement, with some declining as much as 20%. The rule also highlights a policy tension: nearly 900,000 advanced AI chips had been approved for export to Chinese companies, including firms with military connections, in early 2026. Investors will focus on the November effective date, the subsequent 90-day compliance period and enforcement of licensing, reporting and customer-screening requirements.