Nvidia has stopped shipping its licensed Hopper-generation artificial intelligence processors to China, despite receiving permission from Washington to sell some of the chips. The decision follows a quarter in which the company generated $2.4 billion from authorised Chinese sales, highlighting the difficulties of doing business in a market caught between American export controls and Beijing's restrictions.
Interim chief executive Colette Kress confirmed the halt during Nvidia's November 17 earnings call, saying the company was no longer shipping licensed Hopper data-centre products to China and had excluded further sales from its financial outlook.
The announcement came as Nvidia reported record quarterly revenue of $115.8 billion, more than double its sales from a year earlier. Data-centre revenue reached $107.6 billion, demonstrating that the company's global AI business continues expanding rapidly even as its access to Chinese customers narrows.
At the centre of the dispute is the H200, a processor designed for training and running advanced AI models. Although older than Nvidia's latest Rubin-generation hardware, the H200 remains a powerful chip with substantial memory capacity and bandwidth.
Washington had previously restricted exports of advanced processors to China over concerns that the technology could support military applications and sophisticated AI development. The Trump administration subsequently relaxed some restrictions, allowing Nvidia to sell H200 processors to approved Chinese customers under specified conditions.
Those conditions were formalised in a January 2026 Commerce Department rule, which established case-by-case licensing for certain advanced computing products, including the H200. Shipments were subject to security requirements and technical verification before export.
Permission from Washington, however, did not guarantee access to China's market.
Nvidia had already acknowledged in its August regulatory filing that Chinese government restrictions prevented it from selling all the H200 processors covered by its American export licences.
Reports earlier this month suggested that approximately 600,000 H200 processors were being held in Hong Kong while awaiting permission for large-scale imports into mainland China. Nvidia has not publicly verified that inventory figure.
China has also been encouraging greater use of domestically developed semiconductors. Companies including Huawei are expanding their AI hardware portfolios, with new processor architectures and computing systems intended to reduce reliance on foreign suppliers.
For Chinese technology companies, the H200 remains commercially attractive because of Nvidia's established software ecosystem and the processor's ability to support demanding AI workloads. Domestic alternatives are advancing, although differences in computing performance, manufacturing capacity and software compatibility continue to influence purchasing decisions.
The halt comes as Washington tightens oversight of semiconductor exports and moves to restrict China's indirect access to American AI computing through overseas data centres. The latest measures focus on arrangements that could allow restricted customers to use American processors without importing them directly.
Nvidia has not disclosed how long the shipping halt will last or whether deliveries could resume if regulatory conditions change. Kress's announcement also did not establish that the company had permanently withdrawn from China's AI processor market.
For now, Nvidia expects fourth-quarter revenue of approximately $130 billion without assuming any Chinese data-centre computing sales. The company's record earnings have given it room to absorb the interruption, even as one of the world's largest potential AI markets becomes increasingly difficult to serve.