China leads the open-source AI race — but the foundation is still American

Open source in AI is increasingly becoming Chinese at the model layer. But the infrastructure supporting this advancement remains heavily concentrated in one American company: Nvidia.

Victory in models does not end the dispute

It's tempting to look at Chinese leadership in open-source models and conclude that the race has already changed hands. However, this reading is incomplete. In AI, controlling the visible layer — the models — matters a lot. But controlling the invisible layer — chips, toolchains, runtimes, and infrastructure — can matter even more.

What the news truly reveals

China's advancement in open-source AI shows increasing competence in distribution, adoption, and iteration speed. But the fact that Nvidia continues to dominate the underground of AI reveals that the dispute is happening on two different layers.

One layer is the model that gains attention. The other is the infrastructure that allows these models to exist at scale.

Why this matters to the market

This division helps explain why the AI war cannot be read solely in terms of benchmarks or popularity. Even when innovation appears on one side of the world, it can still depend on productive bottlenecks, supply chains, and computational capacity controlled elsewhere.

Conclusion

The Chinese rise in open-source AI is real and relevant. But the article points to a less obvious truth: in AI, winning the application layer does not automatically mean controlling the power layer.