AJP Asia Intelligence Agency

Nvidia Faces Double Blow as Ex-CEO's Return Fails and U.S. Export Controls Tighten

Nvidia is facing two significant setbacks: the failed return of founder Jensen Huang as CEO and the tightening of export controls by the U.S. government. Analysts suggest that revenue related to China accounts for 10% of the total, raising concerns that up to $47 billion in annual revenue could vanish, impacting the entire semiconductor market.


According to the IT industry on November 17, the U.S. Department of Commerce's Bureau of Industry and Security (BIS) announced a provisional final rule on November 16 (local time) to tighten export controls on advanced artificial intelligence (AI) chips and servers.


To export advanced chips to Singapore, Malaysia, Thailand, Indonesia, and Brazil, companies must obtain a license from the U.S. government. Data centers that lease computing power to companies based in China, regardless of their location, are also subject to licensing requirements. Data centers already serving Chinese clients must secure licenses within 90 days of the rule's implementation.


This measure aims to block Chinese companies from circumventing direct chip imports by leasing computing power from third-country data centers. The first revision of export controls follows allegations that a Singaporean company, Megaspeed, linked to Huang, smuggled Nvidia chips into China. On the same day, the possibility of Huang's return was also dismissed, with Nvidia's board stating, "There will be no return for Huang as CEO."


U.S. research firm New Street Research estimates that up to 10% of Nvidia's revenue is directly or indirectly connected to China. This translates to between $3.7 billion and $4.7 billion of the anticipated annual revenue of $41.1 billion. Investment firm Cowen claims that when including indirect sales through third countries, the figure exceeds 20%.


Nvidia's stock closed at $156.14 on November 16, down 4.1% from the previous trading day, marking its lowest closing price since July 2025. The Nasdaq index also fell by 2.49%.


The impact of these controls extends beyond Nvidia alone. The licensing threshold of 12 million TPP (a measure of chip performance) corresponds to approximately 750 Nvidia H100 chips. Commercial AI data centers, which typically require thousands of chips, will effectively all be subject to review.


Southeast Asia is the second-largest market for AI chips after the U.S. Among the 96 facilities tracked globally by AI research organization EpochAI, four in Johor, Malaysia, have a computing capacity equivalent to about 1.44 million H100 chips, second only to the U.S. and 4.3 times more than five facilities in China. Chris McGuire, a researcher at the Council on Foreign Relations (CFR), analyzed that AI chips exported to Singapore, Thailand, and Malaysia this year account for half of the volume leaving the U.S.


The new rules also target the chip production stage. Foundries and packaging companies that manufacture advanced chips based on orders from design firms in these countries will face stricter verification obligations. Since Nvidia's AI chips utilize high-bandwidth memory (HBM) from SK Hynix, Samsung Electronics, and Micron, any delays in the approval process for chips destined for Southeast Asia will also affect memory orders.





* This article has been translated by AI.