When Chinese startup Moonshot unveiled Kimi K3 in July, it did more than release another AI model. It triggered a policy firestorm in Washington. The debate now extends beyond tariffs into the core of how the two superpowers handle artificial intelligence—especially the open-weight models that American companies increasingly rely on. Nvidia’s Jensen Huang and over 20 tech firms signed an open letter warning against premature restrictions, arguing that cutting off Chinese open-source AI would hurt U.S. innovation.

The stakes are high because the interdependence is real. Boston-based venture capitalist Lex Zhao notes that many of his portfolio companies build directly on Chinese models like Kimi and DeepSeek. Meanwhile, the most advanced U.S. open-source model from Thinking Machines Lab acknowledged using those same Chinese tools during development. China is also using AI as a diplomatic lever, hosting back-to-back international events in Shanghai and Chengdu, where U.S. officials attended. The upcoming Trump-Xi summit in September will be the first high-level forum to address AI diplomacy directly.

On the ground, growth is constrained by computing power. Z.ai’s GLM 5.2 coding tools have a three-month waitlist, and Moonshot paused new subscriptions after Kimi K3’s launch due to hardware limits. China is racing to build domestic capacity—Z.ai is constructing a 1-gigawatt data center. Yet recent Pew surveys show more Americans now believe China has pulled ahead in AI development, while roughly half disapprove of both parties’ AI policies. For businesses, the smart move is building flexible systems that can swap between models. As Chinese models improve, ignoring them may become impossible.

Source: CNBC