Chinese AI doesn’t necessarily need to beat American AI; it only needs to break the U.S. AI monopoly on pricing power.

If the Kimi K3 leak is true, I think what it truly threatens isn’t a model ranking, but the pricing power within the U.S. AI narrative.
Now, the valuation of the entire AI industry chain across U.S. stocks is essentially built on one huge premise:
That America’s leading AI companies can maintain an overwhelming advantage long-term.
Model companies can monopolize users.
Cloud providers can monopolize compute power.
Chips, memory, data centers, and electricity can all continuously capture the excess profits along this chain.
That’s why the market is willing to keep giving NVIDIA, MU, Hynix, cloud providers, data centers, and even electricity companies the ability to price the future.
As models like DeepSeek, GLM, and Kimi keep catching up, what they’re essentially telling the market is:
AI doesn’t have to be only one kind of pricing system based on U.S. closed-source models.
If in the future many enterprises and developers find that:
With cheaper Chinese models, they can still meet 80% or even 90% of real needs,
then the U.S. AI excess profit expectations will need to be recalculated.
That’s the most dangerous part.
I even think OpenAI and Anthropic may accelerate their IPOs.
Because right now is the stage when their narrative is strongest, valuations are fullest, and the market is most willing to believe in an “AI monopoly on the future.”
If they keep dragging it out, the gap for Chinese models will keep shrinking, and costs will be even lower.
Once the market realizes that AI isn’t a one-way monopoly game of U.S. closed-source models,
then valuations won’t be a growth story anymore.
They’ll be a repricing.
Whoever controls storage wins, whatever the compute narrative is—ultimately it all relies on monopoly to generate excess profits. If you don’t open a position now, wait for Anthropic to list.