Kimi K3, this AI model, took first place in the global code evaluation! It’s said that its front-end code performance directly surpassed Claude and GPT—its capabilities are pretty strong.

The performance of US chip stocks feels a bit like the “DeepSeek moment” when DeepSeek R1 was released in early 2025. Back then, it also suddenly triggered panic selling.

This time, Kimi K3 is even stronger, and the price is still cheap, and it’s about to open-source its weights. Investors are probably worried: if China’s model is this capable and can also be open-sourced, then in the future, when building AI, do people really not need to keep spending huge money to buy the most expensive chips?

It feels like focusing only on the top-tier chip stocks might not be stable enough. In the future, we should pay more attention to two categories:
First, AI-related companies that can achieve similar results at lower cost.
Second, companies that have advantages in open-source model ecosystems or in deploying AI applications.