Chinese AI giant Zhipu is valued at 1280 times its annual revenue — this is insane even by OpenAI's standards.
The Chinese open-source artificial intelligence developer market is exhibiting an anomaly that defies traditional financial logic. I have analyzed the current situation based on market capitalization and revenue data from key players, and the numbers are, to put it mildly, shocking.
Take Zhipu (Z.ai), the creator of the GLM-5.2 model. After listing on the Hong Kong Stock Exchange on January 8, 2026, and the subsequent release of GLM-5.2, its stock skyrocketed. As of June 22, Zhipu's market capitalization exceeded $118 billion. Now look at its 2025 revenue — it was only about $107 million. Meanwhile, its net loss reached 4.7 billion yuan. A simple calculation gives us a multiple of approximately 1,280 times annual revenue.
For comparison: for Zhipu to trade at "just" 50 times annual revenue, it would need to increase sales to $2.7 billion per year — a 26-fold increase. And for a multiple of 20, it would require $6.9 billion, meaning a 65-fold growth. A similar picture emerges with MiniMax: a market cap of about $23 billion on revenue of $79 million gives a multiple of 290 times.
Why are Chinese AI companies so overvalued?
Against this backdrop, American "labs" look modest. OpenAI, with annual revenue of $25 billion and a private valuation of $852 billion, trades at 34 times annual revenue. Anthropic, with $47 billion in revenue and a valuation of $965 billion, trades at 21 times. But Chinese companies are reaching for the stars. What is the reason?
The key issue, in my view, lies in the revenue structure. A significant portion of Chinese developers' revenue goes to third-party inference providers — such as OpenRouter, Venice, and BaseTen. Users want to work with these models but are unwilling to send data directly to China due to privacy concerns. They turn to intermediaries, which take the lion's share of the profits.
To turn the situation around, Chinese companies will have to prove they do not store user data and offer lower prices than competitors. Doing this is extremely difficult for cultural and social reasons. An alternative scenario is to take equity stakes in American inference providers and sign agreements for early access to top-tier models in exchange for a share of revenue. Then money would flow to developers, albeit as a percentage, but as the overall market volume grows.
My conclusion: The current multiples of Zhipu and MiniMax are a classic bubble, fueled by geopolitical expectations and a shortage of quality public AI assets from China. Fundamentally, such valuations are unjustified, and a correction is inevitable if these companies cannot radically change their monetization model and the trust of international users.