US Requires Licenses for Foreign Data Centers Serving Chinese AI Firms

US Requires Licenses for Foreign Data Centers Serving Chinese AI Firms

Just days before United States President Donald Trump travels to China for high-stakes summit talks with Chinese President Xi Jinping, the US Department of Commerce has unveiled sweeping new semiconductor export controls designed to shut down a major cloud-computing loophole that allowed Chinese artificial intelligence developers to train advanced models using restricted American hardware abroad. Issued on Monday, November 16, 2026, by the Commerce Department's Bureau of Industry and Security (BIS) as an interim final rule, the regulation mandates that foreign data centers providing large-scale AI computing capacity to companies headquartered in China—or other arms-embargoed jurisdictions—must obtain a US export license, with existing data center operators required to secure authorization within a 90-day transition window. Simultaneously, Washington expanded licensing requirements for direct exports of advanced AI semiconductors to five key countries—Singapore, Malaysia, Thailand, Indonesia, and Brazil—after identifying those jurisdictions as carrying a higher-than-normal risk of chip diversion to China. The crackdown arrives amid intense political scrutiny in Washington following the Nvidia–Megaspeed espionage scandal that forced longtime Nvidia CEO Jensen Huang to resign last week, as well as growing pressure on Commerce Secretary Howard Lutnick over export enforcement. Last Friday, US Treasury Secretary Scott Bessent publicly disclosed that four federal agencies favored placing Singapore-based cloud firm Megaspeed—currently under investigation for allegedly diverting Nvidia AI chips to China—onto a US trade blacklist, yet the Commerce Department had delayed acting. Meanwhile, Singapore's government stated on Tuesday that it takes its international reputation seriously, will not condone companies violating foreign export laws, and is engaging US authorities to clarify the scope of the rules, while Chinese AI shares slid in Hong Kong even as domestic Chinese chipmakers rallied on expectations of accelerated self-reliance. For technology executives, semiconductor investors, policy analysts, and readers across New Delhi, Bengaluru, Hyderabad, Mumbai, Lucknow, Singapore, Kuala Lumpur, Beijing, Washington, and Silicon Valley, the new rule redraws the global rules governing AI cloud infrastructure and cross-border compute access.

What the New US AI Data Center and Chip Export Rule Requires

According to the interim final rule published by the Bureau of Industry and Security—which also formally rescinds the Biden administration's January 2025 AI Diffusion Rule—Washington is extending physical semiconductor export controls directly into remote cloud computing:

Regulatory Provision Specific Requirement & Scope Exemptions & Compliance Timeline
Remote Compute Licensing for Foreign Data Centers Foreign data centers exceeding 12 million Total Processing Performance (TPP)—roughly 750 Nvidia H100/H200 GPUs—must obtain a US export license before providing AI compute to entities headquartered in China, Macau, or D:5 embargoed countries Existing data center operators have a 90-day grace period to secure licenses; indirect access via brokers or resellers is also covered
Five-Country Chip Shipment Controls Exporters now require a BIS license to ship advanced AI chips to Singapore, Malaysia, Thailand, Indonesia, and Brazil, flagged for elevated diversion risk US-headquartered cloud companies operating in those five nations are exempt from licensing provided they submit end-user and destination reports
Mandatory Customer Due Diligence Data centers, chip foundries, and exporters must verify whether customers are directly or indirectly owned or controlled by entities in restricted jurisdictions Commercial recommender systems (such as e-commerce and social media algorithms) are exempted from the new licensing mandate

Nvidia–Megaspeed Scandal and Pressure on Commerce Secretary Howard Lutnick

Monday's regulatory action comes on the heels of mounting criticism in Washington that existing export rules failed to stop billions of dollars' worth of advanced Nvidia servers from reaching Chinese AI labs through Southeast Asian data centers. Commerce Secretary Howard Lutnick, who is accompanying President Trump on his Asia trip, has faced sharp questions after reports emerged that he delayed acting on an earlier White House directive targeting suspected diversion networks.

Friction inside the administration spilled into public view on Friday when Treasury Secretary Scott Bessent noted that four US agencies supported blacklisting Singapore-based Megaspeed over alleged illegal transfers of Nvidia hardware to China—an investigation that also triggered scrutiny of Malaysian data center tenant Speedmatrix Sdn Bhd and culminated in Jensen Huang's exit from Nvidia.

Market Reaction Across Asia and Global AI Supply Chains

Following the Commerce Department's announcement, Singapore's Ministry of Trade and Industry affirmed that it is consulting affected firms and seeking "trusted access" talks with Washington. On Asian stock markets Tuesday morning, shares of Chinese tech giants Alibaba (-4.9%) and Tencent (-3.0%) declined alongside broader losses on the Nikkei 225 and Kospi, whereas Chinese semiconductor firms SMIC (+8.7%), Hua Hong (+9.3%), and Cambricon (+5.35%) surged as investors bet Beijing will double down on domestic chipmaking ahead of Trump and Xi's Shenzhen summit.