AI
Nvidia halts H200 shipments to China as new export rules reshape its business
Following Jensen Huang's departure and stricter U.S. export controls, the chip giant abandons its controversial tariff-based arrangement with the country.
TechCrunch — AI
Nvidia is discontinuing sales of its H200 accelerator chips to China, acting CEO Colette Kress revealed on the company's latest earnings call. The move marks a dramatic reversal of the arrangement that had allowed Chinese access to cutting-edge AI hardware under a controversial loophole — a deal that required all shipments to flow through the United States, where the government collected a 25% tariff.
The end of a contentious compromise#
The H200 export pathway was one of outgoing CEO Jensen Huang's signature initiatives, brokered in partnership with the Trump administration. Huang had consistently defended the arrangement, believing it was vital for American tech companies to compete in China's market rather than cede it entirely. When asked about the strategy earlier this year, he argued that "to concede that market for the United States technology industry is a disservice to our country."
However, the deal faced fierce opposition from lawmakers and security experts who wanted to restrict Chinese access to advanced computing power and preserve the U.S. technological edge in artificial intelligence. Huang's recent departure — which involved complications related to an extramarital relationship with a CEO whose company faced smuggling investigations — appears to have shifted the company's stance on this geopolitical issue.
Stricter export rules reshape the landscape#
Nvidia's announcement came the same day the company established a board-level oversight committee focused on national security compliance and global trade risks. More significantly, new export regulations issued Monday now prohibit advanced American chips from reaching foreign data centers that remotely serve Chinese customers.
Investment bank Jefferies predicted these tightened controls could reduce Nvidia's data center revenue by $44 billion to $62 billion annually. Yet Kress dismissed revenue concerns, citing the strong backlog of demand for cutting-edge processors like Vera Rubin.
Broader impact on regional demand#
The long-term fallout may prove more severe. Nvidia faces potential customer losses across Southeast Asia — including Singapore, Malaysia, Thailand, and Indonesia — where demand has been tied to Chinese access and demand. Industry observers estimate that revenue connected to China represents roughly 20% of Nvidia's total business, with the company's Singapore subsidiary alone generating $104 billion in annual revenue in 2025 before Nvidia stopped sharing such disclosures.

