Nvidia H200 China Sales Halted Following New US Export Restrictions

Nvidia officially ended Nvidia H200 China sales under updated US export controls, according to interim CEO Colette Kress during an earnings call.

Advanced Nvidia data center processors running in server racks for artificial intelligence models.
Photo: BoliviaInteligente / Unsplash

Nvidia interim CEO Colette Kress announced that Nvidia H200 China sales have officially ended and shipments are halted under updated US export controls.

Nvidia has officially terminated all Nvidia H200 China sales and suspended hardware shipments following updated US export controls, interim Chief Executive Officer and Chief Financial Officer Colette Kress announced during a financial earnings call on Tuesday, Nov. 17. The policy shift came during the semiconductor maker's first earnings review since co-founder Jensen Huang stepped down, according to ShiftDelete.Net. In response to mounting geopolitical pressures, Nvidia established an executive board committee to oversee national security policies, export control compliance, and global trade exposure.

The decision abruptly closes the door on an export framework that previous company leadership had spent months arranging with Washington.

Framework Collapse and Nvidia H200 China Sales

The discontinued sales mechanism was rooted in an understanding reached between former CEO Jensen Huang and US President Donald Trump. Under that planned structure, processors were not to be delivered directly to Chinese buyers. Instead, shipments were to be routed through the United States, where the American government planned to collect a 25 percent tariff on each transaction. In an April interview with podcaster Dwarkesh Patel, Huang had maintained that abandoning the Chinese market would harm both the US technology industry and broader national interests. That strategy ended abruptly last week when Huang stepped down following an undisclosed relationship with an executive at a technology firm under investigation for chip smuggling.

New Infrastructure Rules Target Overseas Data Centers

Export regulations announced on Monday introduce stricter barriers by prohibiting the sale of advanced American processors to third-country facilities that provide remote computing capacity to Chinese entities. Export control attorney Dan Fisher-Owens told TechCrunch that high-capacity facilities designed for artificial intelligence training must now provide official documentation certifying that they do not serve Chinese clients before purchasing Nvidia silicon. Facilities unable or unwilling to furnish this verification are expected to turn to alternative hardware vendors, which could erode Nvidia's market footprint.

Beijing had already begun turning away shipments before Nvidia confirmed the policy change.

Chinese authorities had refused a substantial share of scheduled H200 shipments prior to the announcement. Jacob Feldgoise, a researcher at Georgetown University's Center for Security and Emerging Technology, noted that Beijing seeks to preserve commercial space for domestic semiconductor manufacturers, even while major domestic artificial intelligence developers express a strong preference for Nvidia hardware. Evidence indicates that all flagship Chinese artificial intelligence models were trained on Nvidia Grace Blackwell processors, which were barred under earlier export controls. Those compute-heavy workloads are widely believed to have run in data centers outside mainland China.

Financial Forecasts and the Vera Rubin Pipeline

International investment bank Jefferies forecasts that the expanded regulatory framework could trigger a $44 billion to $62 billion decline in Nvidia's data center revenue next year. Kress countered that projection during the earnings call, telling investors that strong order backlogs for the upcoming Vera Rubin architecture will bridge the gap and keep overall corporate revenue intact.

Regional exposure across Southeast Asia poses an additional long-term risk.

Market projections indicate that Nvidia could lose enterprise clients across Southeast Asian markets, including Singapore, Malaysia, Thailand, and Indonesia. Industry critics contend that as much as 20 percent of Nvidia's total revenue has derived directly or indirectly from Chinese customer demand. The company's Singapore subsidiary generated $104 billion in annual revenue in 2025 alone, after which Nvidia halted public disclosures of disaggregated regional revenue following regulatory scrutiny.

✍️
Editor's Note: Why It Matters
Nvidia's decision to halt H200 shipments underscores how geopolitical pressures now dictate semiconductor supply chains over commercial ambitions. In TeknoRest's view, the potential $44 billion to $62 billion revenue deficit projected by Jefferies represents a substantial risk to Nvidia's balance sheet. While interim CEO Colette Kress maintains that pent-up demand for Vera Rubin processors will balance this shortfall, mandatory compliance certification for third-country data centers risks permanently fracturing Nvidia's Asian market presence. We believe that as Chinese organizations turn to alternative channels and domestic chip architectures, Nvidia faces an erosion of its global ecosystem dominance that newer hardware cycles cannot easily reverse.

Frequently Asked Questions (FAQ)

Who announced the end of Nvidia H200 sales in China, and when was it disclosed?

Interim CEO and CFO Colette Kress announced the termination of H200 sales to China during a financial earnings call on Tuesday, Nov. 17.

What obligations do the new export rules impose on foreign data centers?

Third-country data centers with capacity for AI model training must provide official documentation proving they do not serve Chinese customers before they can purchase advanced Nvidia processors.

What are the financial estimates regarding Nvidia's data center revenue loss?

Investment bank Jefferies estimates a potential loss of $44 billion to $62 billion in data center revenue next year, though Colette Kress stated that demand for Vera Rubin chips will offset the gap.

Editor: Teknorest Editör Ekibi
🛡️ Primary Source: ShiftDelete.Net | Curated & synthetically analyzed by TeknoRest Editorial Desk with full source attribution (Law No. 5846 Art. 36). Editorial Imprint & Copyright