Summary of Key Points
In just 5 months, Zhipu AI’s market value soared from 51.1 billion to over 700 billion yuan (an increase of more than 13 times), achieving a remarkable turnaround by focusing on the programming industry. However, beneath its success lie three major concerns: a bottleneck in computing power that leads to cost inversion, a long-term lack of profitability, and the potential for significant stock price fluctuations due to the release of restricted shares by major shareholders in July. Additionally, long-term risks such as weak technical barriers and intensifying competition cannot be ignored. The current high valuation is largely driven by faith in AI rather than solid fundamental performance.
1. The Secret Behind the Market Value Surge: Betting on the Profitable Programming Industry
Zhipu initially focused on a B2B approach for government and enterprise clients, earning money through software and hardware solutions leveraging domestic computing power. However, profits were low and competition was fierce (with competitors like iFlytek and SenseTime). The turning point came when programmers began to pay for user-friendly AI programming tools; for example, 37.9% of Claude’s users used it for coding. Zhipu quickly improved its GLM models (from version 4.5 to 5.1), catching up with global leaders in coding and long-context reasoning capabilities, and even dared to raise prices, breaking the traditional “price war” mindset among Chinese companies. As a result, not only did user numbers remain stable, but API usage also increased by 400%. This move positioned Zhipu as the “Chinese version of Anthropic,” which gained success in the U.S. with its Claude Code tool, attracting capital and driving its market value upward.
2. Insufficient Computing Power and Losses: Limited Supply as a Last Resort
Zhipu’s main issue is the inability to keep up with demand for computing power, leading to significant cost inversion. For instance, its Max package costs 469 yuan per month, but the actual cost of the computing resources used by users ranges from 7,000 to 14,000 yuan—meaning the company incurs a loss of 15-30 yuan for every 1 yuan earned. To avoid further losses, Zhipu resorted to “hunger marketing,” offering limited quotas to new domestic users each day, with peak usage (14-18 PM) requiring three times the usual amount of computing power. The CEO stated that despite a 83% increase in API pricing and a 400% rise in usage, demand still outpaces supply due to insufficient computing resources. The company is compensating by scaling down models and dynamically allocating them during peak periods.
3. A Long-Term Lack of Profitability: Questionable Revenue Figures and Weak Technical Barriers
Zhipu’s profitability prospects are uncertain. In 2025, it plans to invest 3.18 billion yuan in research and development, four times its revenue of 724 million yuan, resulting in a decrease in shareholder equity from -395.5 million to -811.1 million yuan. Analysts predict that Zhipu may not start making a profit until 2029, and there are doubts about the accuracy of its revenue figures. The company includes monthly payments from large enterprises in its ARR (Annual Recurring Revenue), but these orders could be canceled due to strategic changes, potentially inflating these figures. Moreover, the technical barriers in the programming industry are not insurmountable; many domestic companies can create decent programming models, and Zhipu does not have a unique advantage. Additionally, unlike Anthropic, which has diverse use cases for its models, Zhipu sells only models, limiting its revenue streams.
4. The July Share Release: A Potential Threat
Zhipu went public in January 2026, with major shareholders and cornerstone investors having a six-month lock-up period, which will expire in July. These shareholders have seen their investments increase by more than tenfold and are highly motivated to sell their shares. The situation is exacerbated by the overall peak in share releases in the second half of the year (over HK$100 billion from cornerstone investors between June and August), potentially putting significant pressure on Zhipu’s stock price. While companies with solid fundamentals can stabilize after share releases, Zhipu’s current lack of profitability raises doubts about whether its high valuation is justified. Even if it achieves a profit of 3.5 billion yuan in 2029 (at a PE ratio of 60), its market value would only be 210 billion yuan, 40% lower than the current 700 billion.
5. Long-Term Risks: Vulnerability to Competition
Zhipu’s position as the leading programming company in China is not secure. Competitors like DeepSeek and Tongyi Qianwen could soon release better models. If Kimi, which focuses on Agent applications, releases a new model with larger parameters, it may attract more attention in the programming industry. Kimi also has a broader range of C-side Agent use cases (such as intelligent assistants) compared to Zhipu, which relies solely on its programming solutions. Furthermore, both Kimi and Jieye Xingchen could go public in the second half of the year, reducing Zhipu’s uniqueness. With rapid model iteration expected over the next three years, today’s advantages may become obsolete soon.
Conclusion
Zhipu AI’s market value surge is a result of its focus on the promising programming industry during the AI boom. However, the high valuation is supported by hopes for future growth rather than current financial performance. Investors should be cautious: if these issues are not resolved, the stock price could experience a significant drop after the share release. Whether Zhipu can realize its potential value in the long term depends on its ability to overcome computing power constraints and establish a sustainable profit model.