Opinion

In what can only be described as the most confident performance of professional gymnastics in recent memory, Nvidia CEO Jensen Huang took to Axios's "Behind the Curtain" interview series this week to deliver a message that would make any geopolitical consultant weep: Washington should not fear Chinese open-source AI models. In fact, Huang argued, the real threat is the "growing campaign to ban them." One might charitably describe this as bold. One might also accurately describe it as the kind of rhetorical position that only becomes possible when you've convinced yourself that your audience isn't paying attention to what you've been doing for the past eighteen months.

Let's establish what we're working with here. Nvidia has built approximately 95% of the global AI infrastructure market—not through superior technology alone, but through ruthless market dominance and supply chain control. The company has explicitly championed, lobbied for, and benefited from U.S. export restrictions on advanced chips to China, restrictions that have become the cornerstone of American AI strategy and leverage in geopolitical negotiations. Huang himself has been a public advocate for these restrictions, appearing before Congress and speaking to the necessity of limiting Chinese access to Nvidia's most advanced silicon. The math is simple: no chips, no competitive Chinese AI. So when Huang now tells the world that Chinese AI models are "excellent" and that banning them would be the real mistake, he's essentially arguing that the weapon he helped forge shouldn't exist—but only after making sure it's pointed in the right direction.

The cognitive dissonance here isn't subtle. Huang's position amounts to saying: "The Chinese models built without our chips are great, so please don't ban them—but also, please keep banning our chips from going to China." This is not a defense of open competition or free markets. This is a defense of a specific outcome: Chinese companies should be allowed to distribute their models freely while remaining technologically dependent on Nvidia's hardware infrastructure. It's the geopolitical equivalent of arguing that your competitor should be allowed to play the game, but only with equipment you've already sold them and won't sell them anymore.

The stated rationale—that banning open-source models represents a greater threat than restricting chip access—inverts the actual power dynamic. Open-source models can be deployed, refined, and improved by anyone, anywhere, with existing infrastructure. That's exactly why they're valuable to China: they represent a way forward that doesn't require Nvidia dominance. But Huang frames restrictions on models as freedom-killing while simultaneously benefiting from restrictions on chips. The logic works only if you believe that Nvidia's supply chain control is invisible—or irrelevant.

What's particularly rich is the timing. Kimi, the Chinese AI model in question, has emerged not despite Nvidia's embargo but *because* of it, forcing Chinese companies to optimize for available hardware and develop their own efficiency advantages. By defending Kimi's right to exist while maintaining his position on chip restrictions, Huang is essentially claiming credit for forcing his competitors to innovate. "Our restrictions made them excellent," he's arguing, "so why restrict their results?"

This is what peak market dominance looks like: the ability to argue that your competitors' success validates your strategy, while your strategy remains completely intact. Huang doesn't want to lose the chip embargo or the competitive advantage it provides. He wants to lose the PR problem of looking like he's afraid of Chinese innovation. In his mind, these aren't contradictory. For the rest of us watching, the contradiction is almost beautiful in its clarity. When your business model depends entirely on control, defending openness is just another form of controlling the narrative.