Nvidia CEO Jensen Huang has publicly defended Chinese artificial intelligence models, urging the United States not to block access to open-source systems that he described as "excellent" and "world class." His remarks, first reported by Axios, place the leader of the world's most valuable chip company directly at odds with a rising tide of Washington pressure to wall off Chinese AI. For the latest AI industry coverage, the comments mark one of the most high-profile industry pushbacks against proposed restrictions on foreign models.
"Excellent Models Should Be Used"
Speaking amid what reporters have dubbed the "Kimi panic" — the market unease that followed Chinese startup Moonshot AI's release of its Kimi K3 model — Huang argued that American companies should be free to adopt strong open-source models regardless of their country of origin. According to TechSpot, Huang said that "open-source models that are excellent should be used."
Tom's Hardware reported that Huang argued American companies should be allowed to use Chinese AI models and dismissed fears of hidden backdoors as "misconceptions." Firstpost quoted the Nvidia chief as saying there was "zero possibility" that the U.S. should fear Chinese AI models, while Crypto Briefing reported that he called them "world class."
The intervention is significant because Nvidia's chips power the vast majority of the world's most advanced AI systems. With the Trump administration weighing restrictions on Chinese open-weight models — and the Treasury reportedly threatening sanctions over allegations that Moonshot "distilled," or copied, technology from rival Anthropic — Huang's stance puts him squarely on the side of open access.
Startups Join the Pushback
Huang is not alone in resisting a crackdown. Seoul Economic Daily reported that U.S. startups are urging the Trump administration to preserve access to Chinese AI models, arguing that blocking them would harm smaller American companies that rely on affordable, openly available foundation models to build products.
Korean outlet Maeil Business Newspaper framed the dynamic as "Jensen Huang and startups push back against Trump regulations," noting that the coalition of resistance extends well beyond one executive. Open-source models from China — including Alibaba's Qwen and Moonshot's Kimi series — have become critical building blocks for developers worldwide precisely because they can be downloaded, modified, and deployed without the per-query costs of proprietary systems like OpenAI's GPT or Anthropic's Claude.
The Kimi K3 Shockwave
The backdrop to Huang's comments is the dramatic market reaction to Moonshot's Kimi K3. The Guardian reported that Moonshot launched Kimi K3 as the first open-source model of its size, advertising it as more customizable than Claude or ChatGPT and judged by analysts as competitive with both leading U.S. models.
The launch sent shockwaves through private markets. Implied valuations for Anthropic's upcoming public debut slumped roughly $232 billion to $1.56 trillion on IG's trading platform between Friday and Tuesday, while OpenAI's implied value dropped about $160 billion to $1.16 trillion, according to The Guardian. Moonshot itself had to temporarily pause new subscriptions after demand for Kimi K3 overwhelmed its compute capacity.
A Conflict of Interest?
Huang's advocacy carries undeniable commercial logic for Nvidia. While U.S. export controls already limit the sale of Nvidia's most advanced data-center chips to China, a flourishing open-source Chinese AI ecosystem still drives enormous demand for inference compute — the less-restricted GPUs that run already-trained models. A broader ban on Chinese models could dampen that demand.
Critics, however, argue that Huang is downplaying genuine national-security concerns. The same week as his comments, the White House's top technology adviser publicly accused Moonshot of stealing from Anthropic, and Bloomberg reported that a U.S. official said Moonshot had accessed banned Nvidia chips. Tom's Hardware noted that Huang characterized concerns about Chinese backdoors as misconceptions — a framing that security hawks in Congress are unlikely to accept.
Open Source as a Global Commodity
The tension Huang is navigating reflects a deeper structural shift. For most of software history, the most valuable code was proprietary and closely guarded. Open-source AI has inverted that assumption: some of the most capable models in the world are now freely downloadable, modifiable, and redistributable. That openness has democratized access for researchers and startups, but it has also made it nearly impossible to contain the spread of powerful systems once they are released.
For Nvidia, which sells the GPUs that run both proprietary and open-source models, the calculus is straightforward. More models in circulation — whether American or Chinese — means more demand for inference hardware. A U.S. crackdown that drove Chinese open-source models underground would not necessarily eliminate them; it would simply redirect the compute spending toward non-U.S. suppliers. Huang's argument, in effect, is that the United States gains more by staying plugged into the global open-source ecosystem than by trying to wall it off.
What Comes Next
The debate is converging on a single question: should the United States treat powerful open-weight AI models as freely tradable software, or as controlled technology subject to sanctions? Startup Fortune reported that Congress is moving to make "AI model distillation" — the technique of training a cheaper model to mimic a more expensive one — a sanctionable offense, directly targeting the practice of which Moonshot stands accused.
For now, Nvidia's CEO has drawn a clear line. "Open-source models that are excellent should be used," Huang insisted — a simple principle with enormous commercial and geopolitical consequences.
Stay Ahead of AI
The race between American and Chinese AI labs is reshaping markets, policy, and the global balance of technology. Follow breaking AI news as the open-source debate intensifies.
Read more AI news →
