Nvidia H200 China Exports Stalled by Dual US-China Restrictions

Nvidia's long-anticipated export agreement for its H200 artificial intelligence chip to Chinese technology firms remains stalled, not because of explicit US government restrictions but rather due to conflicting regulatory demands from both Washington and Beijing.

Diplomatic Overtures Yield No Movement

In the wake of President Trump's recent summit with Chinese President Xi Jinping in Beijing, expectations were high that the chip export could proceed following last-minute diplomatic efforts, including the unexpected inclusion of Nvidia CEO Jensen Huang in the US delegation. Nevertheless, no shipments of the Nvidia H200 to China have been made since Trump authorized the exports in December 2025. US Trade Representative Jamieson Greer confirmed that semiconductor export controls were not part of the summit's formal agenda.

Licences Approved but Deliveries Blocked

Approximately ten major Chinese firms—including Alibaba, Tencent, ByteDance, and JD.com—currently hold US export licences allowing them to import up to 75,000 H200 units per company. Distributors Lenovo and Foxconn are also authorized to facilitate shipments. However, the chips remain undelivered because the Chinese government has not given domestic companies permission to take delivery, despite US regulatory approval.

Conflicting National Policies

The impasse results from regulatory requirements that are fundamentally incompatible. US export rules mandate that H200 chips sold to China must be deployed exclusively within the country. In contrast, Chinese authorities have instructed local technology firms to use imported Nvidia chips only for overseas operations, while reserving domestic deployment for locally produced hardware. As a result, chips cleared for export to China cannot legally be used the way Beijing prefers, and Chinese firms are effectively blocked from accepting shipments under current conditions.

This regulatory standoff serves a strategic purpose for China, which is actively reducing its dependence on US semiconductors. Commerce Secretary Howard Lutnick recently testified that Chinese companies are redirecting investment toward domestic technology suppliers, including Huawei. The Chinese State Council has also launched a supply-chain security review aimed at further reducing reliance on US-sourced chips.

Rise of Domestic Players

While diplomatic discussions failed to alter the export status, the summit period coincided with visible advancements among Chinese semiconductor and AI firms. DeepSeek confirmed its latest AI model had been optimised to run on Huawei processors. Executives from Tencent and Alibaba reported ongoing progression in the domestic supply of GPUs and proprietary chip deployment. In April, DeepSeek V4 became the first major Chinese AI model to use Huawei’s Ascend chips for model training, not just inference, signalling increased confidence in domestic hardware.

These shifts have led to a pronounced decline in Nvidia's China revenues, now representing just 5% of company sales, down sharply from over 20% before the US strengthened export controls. Nvidia’s own guidance for the current quarter assumes zero revenue from China.

Industry Implications

The AI industry's trajectory in China is being increasingly dictated by government mandate rather than market or technical benchmarks. With authorities encouraging leading platforms to standardize on Huawei’s compute stack, the competitiveness of Nvidia hardware is becoming less relevant in the world’s second-largest AI market. Early results, such as those from DeepSeek V4, indicate that for certain applications—especially inference, or the deployment phase of AI models—the performance gap between Huawei and Nvidia hardware may be narrowing.

A Structural Roadblock

Despite public assertions from US and Chinese officials hinting at possible compromise, the deal remains immobilized by deeper strategic considerations reflecting both nations’ long-term priorities. US policy clears the Nvidia H200 for China, but only on terms Beijing refuses to accept, while China moves to reinforce self-reliance in its critical technology infrastructure.

Reference: artificialintelligence-news.com