
Nvidia Halts H200 Chip Sales in China After Jensen Huang's Ouster
By Ali Al-Montaser — Senior AI Technical Editor
Nvidia officially stops selling H200 chips in China through interim CEO Colette Kress, ending a 25% tariff loophole. Jefferies expects $44–62 billion in data center revenue loss next year, while Singapore branch revenue hit $104 billion in 2025, with up to 20% of revenue dependent on China-linked demand.
Executive Overview
Nvidia has officially halted sales of its H200 chips in China, interim CEO Colette Kress announced Tuesday, closing a 25% tariff loophole that had allowed shipments through the United States. The decision follows last week's ouster of founder and CEO Jensen Huang and coincides with the formation of a board-level committee to oversee national security and export control compliance. Jefferies estimates the new restrictions could cut $44 billion to $62 billion from Nvidia's data center revenue next year, while the company's Singapore branch generated $104 billion in 2025, with up to 20% of total revenue tied to China-linked demand.
📊 Official Technical Specifications & Data Sheet
| Technical Axis | Confirmed Official Data |
|---|---|
| 💰 Pricing & Usage Cost | 25% tariff on H200 shipments through the United States (under previous arrangement); no token prices associated with this announcement. |
| 🌐 Platforms & Immediate Availability | H200 chips were shipped to China via the United States; the decision halts this route. Advanced chips like Vera Rubin and Grace Blackwell are subject to US export controls. |
| ⚡ Performance & Speed Benchmarks | No new performance figures in the announcement; H200, Vera Rubin, and Grace Blackwell are referenced as advanced AI chips. |
| 🛡️ Security & Breach Resistance | Formation of a board-level committee to oversee national security policies, export control compliance, and global trade risks. |
| 🧠 Context Window | Not applicable to this announcement (pertains to chips, not language models). |
| 🌍 Arabic Language & Regional Support | Not applicable; the announcement concerns chip export controls to China, not language support. |
Deep-Dive Features & Architecture
The announcement came during Nvidia's first earnings call after the departure of founder and CEO Jensen Huang last week. Huang had brokered a deal to sell H200 chips in China with President Donald Trump, which required shipping the chips through the United States to collect a 25% tariff. This route angered China hawks who wanted to deny Chinese labs computing power, while Huang advocated for competition in the Chinese market.
Nvidia's decision coincided with new rules issued Monday that prohibit foreign data centers serving Chinese customers remotely from using advanced US chips. Jefferies expects these rules to reduce Nvidia's data center revenue by $44 billion to $62 billion next year, but Kress said that accumulated demand for advanced chips like Vera Rubin means the rules will not affect the company's overall revenue.
In the long term, Nvidia may face losing potential customers in Singapore, Malaysia, Thailand, and Indonesia. Critics emphasize that up to 20% of the company's revenue comes from China-linked demand, and Singapore branch revenue reached $104 billion in 2025 before Nvidia stopped disclosing those figures.
Benchmark & Competitive Performance
Even before the decision to halt H200 shipments, China had largely refused to accept them, preferring to generate demand for its domestic chip industry, according to Jacob Feldgoise, a researcher at Georgetown University's Center for Security and Emerging Technology. Meanwhile, leading Chinese labs are finding ways to access Nvidia chips from elsewhere; evidence suggests that all leading Chinese models were trained on Grace Blackwell chips, which are banned from export to China. Under the new rules, large data centers wishing to purchase Nvidia chips must certify that they do not serve any Chinese customers, or they may turn to alternative chip options, further eroding Nvidia's market share.
Industry Impact & Enterprise Adoption
This announcement does not directly concern token prices or Arabic language support, but it affects the global computing infrastructure that AI models rely on. With tightening export controls, data centers in the Arab region may need to verify that they do not serve Chinese customers to continue purchasing advanced Nvidia chips, potentially raising development costs for Arab startups relying on cloud inference. Diversifying chip sources could open opportunities for regional data center partnerships, but may slow access to the latest capabilities.
Conclusion
Nvidia's halt of H200 sales in China marks a significant shift in its China strategy following Jensen Huang's departure. The move, driven by new export rules and national security concerns, is expected to cost the company billions in data center revenue next year. While Nvidia bets on demand for its advanced Vera Rubin chips to offset losses, the long-term impact on its market share and global AI infrastructure remains uncertain.
Media Source: TechCrunch AI | Fact Verification & Analysis: AI Tools Oasis
Frequently Asked Questions
Interim CEO Colette Kress announced on Tuesday that Nvidia will no longer sell H200 chips in China, ending its participation in a loophole that allowed shipments through the United States with a 25% tariff.
Jefferies expects the new rules to reduce Nvidia's data center revenue by between $44 billion and $62 billion next year.
Nvidia's Singapore branch revenue reached $104 billion in 2025 before the company stopped disclosing those figures.
Critics argue that up to 20% of Nvidia's revenue comes from China-linked demand.
Evidence suggests that all leading Chinese AI models were trained on Nvidia Grace Blackwell chips, which are banned from export to China.

AI Tools Oasis Team
Bringing you the latest news and analysis in the world of Artificial Intelligence with accuracy and credibility. Follow us for all updates.


