Nvidia CEO Jensen Huang has sharply criticized U.S. export restrictions on artificial intelligence chips to China, calling them a “failure” that cost American companies billions in lost sales. Speaking at Computex in Taipei, Huang said the Biden administration’s AI diffusion rule, which aimed to limit China’s access to advanced semiconductors by dividing countries into export tiers, was based on flawed assumptions. He noted that despite the restrictions, AI research and development continue unabated in China, which is now investing heavily in building a self-reliant chip supply chain.
Huang emphasized that the U.S. ban drove Chinese firms to turn to local chipmakers like Huawei, intensifying competition in the Chinese market. Nvidia’s share in China has fallen from 95% to 50% since the start of the Biden administration, with the company projecting that China’s AI market could reach $50 billion next year. He lauded former President Donald Trump’s efforts to revise the policy, arguing that global cooperation and licensing would better serve American trade interests than unilateral restrictions.
The export curbs have also had a significant financial impact on Nvidia. The company recently said it would take $5.5 billion in charges after restrictions blocked sales of its H20 chip to China, with Huang estimating a $15 billion revenue hit. In response, Nvidia is now developing a compliant version of its new Blackwell AI chip with reduced memory speed to meet U.S. export rules while maintaining competitiveness in global markets.
China has pushed back strongly against the U.S. restrictions, demanding that Washington reverse what it calls “discriminatory” actions. The Chinese Commerce Ministry criticized the U.S. for undermining bilateral trade agreements and warned of resolute retaliatory measures if the curbs persist. As the Biden administration considers replacing the current tiered export framework with a global licensing system, Nvidia’s stance underscores the growing tension between national security concerns and economic imperatives in the high-stakes AI sector.