Huang: Export Controls Risk Leadership

Nvidia chief executive Jensen Huang said tighter U.S. chip curbs on China could weaken, not protect, America’s position in artificial intelligence. (enterpriseai.economictimes.indiatimes.com) Huang made the case in a Dwarkesh Podcast episode published April 15, saying Washington should use “more balanced” rules instead of “giving up a vast part of the world’s market.” (dwarkesh.com, enterpriseai.economictimes.indiatimes.com) His warning landed after Nvidia disclosed on April 9, 2025 that the U.S. government would require a license for exports of its H20 chip to China, Hong Kong, Macau, and some related buyers, and said the restriction could force about $5.5 billion in charges. (sec.gov) The policy backdrop has shifted more than once. The Commerce Department’s January 15, 2025 “Framework for Artificial Intelligence Diffusion” expanded controls on advanced computing chips and some artificial intelligence model weights, then on January 13, 2026 the Bureau of Industry and Security said it would review some high-end chip exports to China case by case. (federalregister.gov, bis.gov) Huang’s argument is about software standards as much as hardware sales. He said Chinese developers pushed off Nvidia’s chips could move from Nvidia’s CUDA programming platform to Huawei’s CANN stack, shifting the tools, code, and habits that lock in a computing ecosystem. (enterpriseai.economictimes.indiatimes.com, dwarkesh.com) He tied that risk to China’s research base, saying recent gains in artificial intelligence have come from algorithms and software as well as raw chip performance. In the same interview, he pointed to DeepSeek and said a future in which that work runs on Huawei first would be “a horrible outcome” for the United States. (enterpriseai.economictimes.indiatimes.com, dwarkesh.com, thenextweb.com) The U.S. government has framed the same controls in national-security terms. Nvidia’s April 2025 filing said officials told the company the H20 license rule was meant to address the risk that the chips could be used in, or diverted to, a supercomputer in China. (sec.gov) Nvidia’s own filings show why the China question stays central even after earlier curbs. In its annual report for the year ended January 26, 2025, the company said data-center revenue in China grew in fiscal 2025, though it remained well below the levels seen before export controls tightened in October 2023. (sec.gov, publicnow.com) That leaves buyers, cloud companies, and model developers planning around two risks at once: whether U.S. licenses will tighten again, and whether China will keep building alternatives fast enough to reduce Nvidia’s leverage. Huang’s point was that export controls can shape whose tools become the default, not just which chips ship this quarter. (enterpriseai.economictimes.indiatimes.com, bis.gov, federalregister.gov)