Key insights
- The author argues that Chinese AI models, offering near-frontier capabilities at significantly lower costs, threaten the valuation of US AI companies. The expectation of future monopoly margins is undermined by this competition, leading to margin compression and price cuts. This situation is prompting US companies to seek government support and bundle services, signaling a potential shift in AI leadership.
imo - The AI race is not going well and might be over.
China doesn’t need to “beat” OpenAI feature-for-feature tomorrow. They just need models that are 90-95% as good at 10-20% of the cost. That completely nukes the economics these US companies were valued on.
Look at chinese models: Kimi, Qwen, GLM, MiniMax, DeepSeek
They’re absurdly capable, especially for coding and agent workflows. Cursor literally based their composer 2.0 model on chinese Kimi 2.5 - because developers care about output and price, not patriotic fanfiction from SF VCs. And I would argue - that their best IP - a composer 2.0 model and it apparently based on chinese model
And here’s the fatal problem nobody wants to admit:
US AI valuations were built on the assumption that after initial user base acquisition they can raise prices. The entire bubble depends on eventually charging enterprise customers massive premiums.
But you cannot raise prices when China is shipping near-frontier models for a fraction of the cost.
That destroys the “future monopoly margins” story.
Anthropic publishing bleating pieces like “2028 AI Leadership” reads less like confidence and more like a cry for help - like positioning for government support because they can see the wall coming. Suddenly everyone is talking about “national security,” “strategic leadership,” and “infrastructure subsidies.”
Why? Because the market is realizing open-weight and Chinese competition are compressing margins into oblivion.
And notice the timing:
* Microsoft already cut Copilot pricing/tiering.
* Anthropic adjusted tiers.
* Everyone is racing to bundle and discount before the commodity wave fully hits.
That’s not what dominance looks like.
That’s what margin panic looks like.
The US still has incredible research talent. But economically? The moat is evaporating in real time.
Software always trends toward cheaper, faster, more available competition. AI won’t magically violate that law just because Silicon Valley wants trillion-dollar valuations.
the takeaway - I would not invest into US ai companies.
edit:
- China has access to almost free electric and we dont * China has firewall and we dont (thus they can use all our content, and we cannot user theirs )