China Restricts Nvidia’s H20 AI Chip Sales Following U.S. Official’s Remarks Chinese authorities have moved to limit the sale of Nvidia’s H20 artificial intelligence (AI) chip within the country. This decision follows comments made by U.S. Commerce Secretary Howard Lutnick, which Chinese officials found “insulting.” Lutnick stated that the United States sells downgraded chips to […]
Chinese authorities have moved to limit the sale of Nvidia’s H20 artificial intelligence (AI) chip within the country. This decision follows comments made by U.S. Commerce Secretary Howard Lutnick, which Chinese officials found “insulting.” Lutnick stated that the United States sells downgraded chips to China to make Chinese developers reliant on American technology. In response, major Chinese regulatory bodies, including the Cyberspace Administration of China (CAC), the National Development and Reform Commission (NDRC), and the Ministry of Industry and Information Technology (MIIT), have advised domestic technology firms such as Alibaba and ByteDance to halt or reduce orders of the H20 chip. These regulators cited national security concerns and urged a shift toward domestically produced chips.
Prior to Lutnick’s remarks, Nvidia’s H20 chip had garnered significant interest in China, and Nvidia CEO Jensen Huang had received a positive reception in Beijing. However, the recent regulatory guidance reflects China’s broader strategy to achieve technological self-sufficiency, especially in critical sectors like AI. Despite internal differences within the Chinese government regarding Nvidia’s role, the current stance underscores a commitment to reducing dependence on foreign technology. U.S. export controls have already restricted Nvidia from selling its high-end Blackwell chips to China, though a downgraded version may be permitted in the future. In the interim, Chinese companies are being encouraged to adopt local alternatives, such as those produced by Huawei and Cambricon. The regulatory guidance remains informal at this stage, leaving room for potential policy adjustments amid ongoing U.S.–China trade tensions.
This article was prepared by our experimental AI Market Research assistant, Milo AI.
John O'Connor is the founder and principal engineer of Web Lifter, a Brisbane software studio building custom software, AI systems, and structured data for Australian SMBs. He has spent over eight years shipping production AI and backend systems, and writes about what actually holds up once the demos are over. Everything published here is drawn from systems running in production for real clients.