Nvidia CEO Jensen Huang told US companies to use Chinese AI models, calling them "world class" after Moonshot AI's Kimi K3 triggered a 20% semiconductor stock selloff.
Nvidia CEO Jensen Huang told US companies to use Chinese AI models, calling them "world class" after Moonshot AI's Kimi K3 triggered a 20% semiconductor stock selloff.

Jensen Huang urged US companies to "absolutely" use Chinese open-source AI models, pushing back against Washington's push to ban them as Moonshot AI's Kimi K3 triggered a 20% selloff in semiconductor stocks.
"China's open-source AI ecosystem is thriving and globally competitive," Huang, chief executive officer of Nvidia, said at the China International Supply Chain Expo. He specifically cited models from DeepSeek, Alibaba, Tencent, MiniMax and Baidu as evidence of the country's AI capabilities.
The comments came after Moonshot AI's Kimi K3 — a 2.8 trillion-parameter model with a 1 million-token context window — topped the Frontend Code Arena leaderboard and scored highly on GDPval-AA v2 tests. The release erased more than 20% from semiconductor stocks from their June 2026 peaks and briefly knocked Nvidia from its position as the world's most valuable company.
Huang drew a distinction between hardware and software restrictions, endorsing limits on exports of Nvidia's most advanced Blackwell and Rubin architectures to China while arguing against banning Chinese AI models themselves. Open-weight models are "better for safety and collaboration" than closed proprietary systems, he said, putting him at odds with Trump administration officials and US AI labs lobbying for a full ban.
The stance carries strategic weight for Nvidia, which generates roughly a fifth of its revenue from China through sales of lower-spec chips compliant with US export controls. A fragmented AI market — where certain models are banned in certain jurisdictions — could accelerate demand for censorship-resistant infrastructure, benefiting decentralized compute networks while complicating Nvidia's China market access.
Nvidia shares have recovered some ground since the Kimi K3 selloff but remain below their June 2026 highs. The company trades at roughly 35x forward earnings, a premium that reflects its dominant position in AI training chips but also embeds risk from both Chinese competition and export policy uncertainty. Huang's public embrace of Chinese AI may signal confidence that Nvidia's hardware moat — built on its CUDA software ecosystem and next-generation Rubin architecture — remains intact regardless of which models win the software race.
This article is for informational purposes only and does not constitute investment advice.