Nvidia's Jensen Huang Defends Chinese AI Amid "Kimi Panic": "Free AI Should Be Great for Chips"

 Nvidia's Jensen Huang Defends Chinese AI Amid "Kimi Panic": "Free AI Should Be Great for Chips"

**The CEO of the world's most valuable chipmaker just made a case that Wall Street and Washington have the AI threat exactly backward—and it could reshape the debate over Chinese open-source models.**

---

## Introduction: The "Kimi Shock" That Changed Everything

Just a week ago, the AI world was rocked by a familiar pattern: a Chinese AI model arrived that seemed to do more with less. Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weight model, topped front-end coding benchmarks, matched the best from OpenAI and Anthropic, and cost 40% less. The market reacted with a panic that echoed the DeepSeek selloff of January 2025. Semiconductor stocks dropped more than 20% from their June peaks. Nvidia itself briefly lost its crown as the world's most valuable company.

But this time, the reaction came with a twist. Instead of defending the fortress of American AI supremacy, Nvidia's CEO Jensen Huang did something unexpected: he praised the competition.

In an exclusive interview with Axios from a new Nvidia plant in Texas, Huang said American companies should "absolutely" be allowed to use Chinese open-source AI models. "These Chinese models are excellent," he told Axios' Mike Allen. "Open-source models that are excellent should be used".

The remarks placed Huang directly at odds with Washington—where Treasury Secretary Scott Bessent had just warned of potential sanctions on Chinese models for alleged intellectual property theft—and with his own largest customers, OpenAI and Anthropic, who have lobbied aggressively to restrict access to Chinese AI.

But Huang's argument wasn't just a contrarian take. It was a fully formed strategic vision about how the AI economy actually works.

---

## The "Counterintuitive" Logic: Why Cheaper AI Is Better for Nvidia

Huang's argument is deceptively simple: cheaper, more accessible AI models create more AI users, and more AI users create more demand for the chips, data centers, and infrastructure that Nvidia sells.

"The market misunderstood the impact of DeepSeek the first time," Huang said, adding that Wall Street has "misunderstood the impact of Kimi again this time".

His logic echoes the "Jevons Paradox"—the economic observation that efficiency gains often lead to increased consumption, not less. In Huang's view, open-weight models like Kimi K3 don't threaten Nvidia's business; they expand the addressable market. "Free AI should be great for hardware," he said. "Free AI should be great for chips. Free AI should be great for data centers".

The technical analysis supports his case. SemiAnalysis noted that while Kimi K3 uses an innovative architecture to reduce computational load, its massive parameter count and massive model architecture require at least 56 chips just to run inference. To run it efficiently, you need precisely the kind of high-end infrastructure that Nvidia sells—and more of it, because the reduced KV cache actually increases cross-chip communication bandwidth demands, deepening the need for Nvidia's proprietary NVLink interconnect technology.

In other words: the "efficient" Chinese model still requires a lot of Nvidia hardware.

---

## Security vs. Openness: Huang's Uncomfortable Truth

Huang also took direct aim at the security rationale for restricting Chinese AI. Critics have warned that open-weight Chinese models could contain "backdoors" or be used to steal U.S. intellectual property. But Huang argues the opposite: openness actually makes AI more secure.

He noted that companies can run downloaded models in secure "sandboxes," and that outside researchers can inspect open-weight models, expose vulnerabilities, and build defenses in ways they can't with closed systems. "If everything just becomes one single model, one single point of attack, one single source of failure, I think the world is much, much more vulnerable," he said.

He rejected the idea that downloading a Chinese model creates a "backdoor" to Beijing. "Companies can customize them and control their access inside secure sandboxes".

Huang also addressed allegations that Chinese models are "distilling" U.S. models—training their own systems on the outputs of American frontier AI. "Distillation, learning from AI, learning from other sources of knowledge, is fundamental to intelligence," he told Axios. He said companies should still face consequences for violating privacy or contracts, but his broader message was clear: target the misconduct, not the models.

---

## The Bigger Picture: A Battle Over AI's Future

Huang's comments came just hours after Treasury Secretary Scott Bessent told Fox Business that the administration is examining Chinese AI models for "watermarks" of stolen U.S. intellectual property and considering sanctions in response. "If we see ... that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent said.

But some AI experts have pushed back on the distillation narrative. "Distillation is a practice that everyone is doing, including companies in the US," said Hugging Face CEO Clem Delangue. He called it "a very small factor".

Huang also called for Anthropic to make its restricted Claude Mythos model available to "everyone," rather than limiting access to a few approved U.S. companies. "Just because Mythos is not available, open models are available anyhow. So I think: Let Anthropic run," he said. "Holding Anthropic back is not in the benefit of the United States".

---

## Frequently Asked Questions

**Q: What is Kimi K3?**

A: Kimi K3 is a 2.8-trillion-parameter, open-weight AI model from Beijing-based Moonshot AI. It topped front-end coding benchmarks, performed competitively against OpenAI and Anthropic's best models, and costs 40% less.

**Q: Why did Huang defend Chinese AI models?**

A: Huang argues that cheaper open-source models expand the AI market, creating more demand for the chips, data centers, and infrastructure that Nvidia sells. "Free AI should be great for chips," he said.

**Q: Does this conflict with Nvidia's national security stance?**

A: Huang has previously said "national security comes first" and supports restrictions on exporting Nvidia's most advanced chips to China. But he argues the software layer—AI models themselves—should remain open.

**Q: What did Treasury Secretary Bessent say?**

A: Bessent warned that the administration is examining Chinese AI models for evidence of stolen U.S. intellectual property and is considering sanctions. He cited "watermarks" of U.S. models found in Chinese systems.

**Q: Are Chinese models actually a threat?**

A: Huang says no. He dismissed the idea that "China runs U.S. companies off the road" as "zero possibility" and argued that U.S. and China will coexist in the AI race.

--Read more-

## Conclusion: A Counterintuitive Bet

Jensen Huang's decision to publicly defend Chinese AI—at a moment when Nvidia's stock was under pressure and Washington was threatening sanctions—reflects a strategic calculation that the AI industry's future lies in expansion, not restriction.

He sees a world where cheaper, open models bring AI to billions of people, and where the hardware companies that power that expansion profit accordingly. His argument is that fear of Chinese competition is misplaced, that open source makes AI safer, and that the real threat isn't China's models but the campaign to block them.

Whether he's right or wrong, his position has already shaped the debate. And for a company that makes the chips that run the AI world, that's a position worth defending.