Chinese AI giant Zhipu valued at 1280 times annual revenue: bubble or new reality?
The artificial intelligence market continues to surprise with its volatility, and the latest example from China raises questions about a fundamental revaluation of the sector. The developer of the GLM-5.2 model, company Zhipu (Z.ai), which listed on the Hong Kong Stock Exchange on January 8, 2026, is showing anomalous multiples that are tens of times higher than those of industry leaders like OpenAI and Anthropic.
At its peak on June 22, Zhipu's market capitalization exceeded $118 billion, while the company's revenue for 2025 was only about $107 million. The net loss for the same period reached 4.7 billion yuan. Thus, the company is trading at a multiple of about 1,280 times annual revenue. In my professional opinion, this is a classic sign of an overheated market, where investors are betting not on current financial performance but on future exponential growth.
Why are Chinese AI companies so expensive?
For Zhipu to justify its current valuation, it needs to achieve a real feat. According to my calculations, to trade at 50 times annual revenue, it would need to increase sales to $2.7 billion per year — 26 times the current level. For a multiple of 20, it would require $6.9 billion, or growth of 65 times. A similar situation applies to another Chinese player, MiniMax: with a market capitalization of about $23 billion and revenue of $79 million, the multiple reaches 290.
For comparison, American "labs" look much more modest. OpenAI, with annual revenue of $25 billion and a latest private valuation of $852 billion, trades at about 34 times annual revenue. Anthropic, with a result of $47 billion and a valuation of $965 billion, trades at about 21 times. The gap is enormous and cannot be explained solely by differences in growth rates.
What is the root of the problem?
The key reason for this imbalance, in my view, lies in the revenue structure of Chinese AI companies. A significant portion of their 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, so they turn to intermediaries.
To turn the situation around, Chinese developers will have to prove that they do not store user data and offer lower prices than competitors. However, given cultural and social characteristics, this is an extremely difficult task.
An alternative scenario, which I consider more realistic, is for Chinese companies to take equity stakes in American inference providers in exchange for a share of revenue. In this case, money would begin flowing to model developers through a percentage of the intermediaries' turnover, allowing them to monetize the growing market without direct access to end users.
My conclusion: The current valuation of Zhipu and other Chinese AI companies is not so much a reflection of their current business performance as it is an advance payment for future dominance in the open-source model segment. However, without solving the data trust issue and building effective distribution through Western partners, this bubble risks bursting.