Summary of Key Points
On September 1st, Zhipu's Hong Kong stock price first fell and then rose (opening lower by 4.4% before the market session and rising by 1.51% during the session). This movement was driven by the company's semi-annual report for 2026, which was released the previous day: revenue amounted to 954 million yuan (an increase of nearly four times compared to the same period in 2025), but the company incurred a loss of 2.072 billion yuan (a 12% reduction). The main issue is that the revenue fell significantly short of institutional expectations. Institutions had forecasted annual revenue to range from 5 billion to 8.5 billion yuan, yet Zhipu only achieved 954 million yuan in the first half of the year, meaning it needs to generate an additional 4 billion to 7.5 billion yuan in the second half to meet these targets, which has raised doubts among investors. However, China Merchants Bank International (CMBI) continued to hold its shares, citing Zhipu's 1.6 billion yuan in annual recurring revenue (ARR) in August and the company's goal of reaching 2.4 billion yuan by the end of the year. The management explained that Zhipu's business model is shifting from focusing on selling tools to providing tangible results. While the company has impressive user numbers and computing power capabilities, the industry competition is fierce, as there is no significant technological advantage between models, and the market is moving away from price-based competition towards a focus on intelligent capabilities.
Detailed Analysis
1. Revenue Falls Short of Expectations: Why Were Institutions So Aggressive?
Institutions had previously projected Zhipu's annual revenue to be as high as 5 billion to 8.5 billion yuan (with Goldman Sachs even estimating 8.5 billion yuan). With only 954 million yuan achieved in the first half, Zhipu would need to generate an additional 4 billion to 7.5 billion yuan in the second half, which seems almost impossible, leading to widespread skepticism in the market. The reason for such high expectations was the company's potential to replicate the explosive growth of Anthropic, another AI company. Anthropic managed to increase its annual recurring revenue from $1 billion to $5 billion in just half a year. Institutions used a valuation method based on the market sales rate (the idea that higher revenue makes a company more valuable), but since Zhipu failed to meet these targets, doubts arose.
2. Zhipu's Business Model Evolution: From Selling Tools to Providing Solutions
The management stated in a call that their revenue generation strategy is evolving:
- Early Stage: Zhipu focused on selling model deployment solutions, where customers had to integrate the models into their own systems (similar to selling a hammer and expecting them to use it to drive nails).
- 2025: The company shifted to offering coding services, providing API interfaces and subscription-based programming solutions, reducing the need for customized model deployments.
- This Year: Zhipu has entered the "Agent" phase, where models can directly perform tasks (for example, generating reports without the need for users to adjust parameters).
These changes are supported by solid financial data: the number of users has increased by 144% since the beginning of the year, with daily active paid users rising by 603%. The top ten customers are using Zhipu's services more frequently (daily calls have increased by 98 times), and the computing efficiency has improved by 14 times, making the models faster and more cost-effective.
3. Fierce Industry Competition: Is Zhipu at a Disadvantage?
Zhipu aims to become the Chinese equivalent of Anthropic, but its model capabilities do not yet offer a significant competitive edge. For example, if other companies use 4G technology, Zhipu also has to use 4G, without any clear technological advantages. After the release of Kimi K3 in July, JPMorgan Chase reduced Zhipu's valuation multiple from 30 to 20 times and downgraded its status from "leader in domestic AI models" to "one of the cutting-edge AI laboratories in China." The competition is rapidly evolving, with new models such as Kimi K3, DeepSeek-V4-Flash, and Qwen3.8-Flash emerging, giving developers more choices. Companies without a competitive advantage are forced to lower their prices, losing control over pricing.
4. Changing Industry Trends: From Price Competition to Intelligence Competitiveness
The AI industry is shifting from a focus on low prices to a emphasis on intelligent capabilities. Morgan Stanley notes three key changes:
- Rational Pricing: Companies are no longer aggressively cutting prices to attract customers.
- Tighter Open Source Policies: Models that were previously available for free are now being made more restricted.
- Advanced Model Complexity: Models are becoming more complex and powerful.
For Zhipu, this shift is beneficial, as models with similar intelligence levels will become cheaper, but those that can solve new problems more effectively or have higher success rates will retain higher prices. However, if Zhipu cannot create a significant technological advantage, it will be at a disadvantage in this highly competitive market.
5. Stock Price Volatility: Market Sentiment Swings
- Initial Drop: Investors were disappointed by the lower-than-expected revenue.
- Midday Recovery: CMBI's announcement of 1.6 billion yuan in ARR in August and the management's goal of 2.4 billion yuan by the end of the year boosted market sentiment, as these figures indicated potential growth. Additionally, positive data on user growth and improved computing power strengthened the belief that Zhipu has a solid foundation.
In summary, the market is concerned that Zhipu may not meet its institutional targets, but it also recognizes the effectiveness of its business model and user growth. As a result, the stock price experienced a sharp decline followed by a recovery.
Final Conclusion
Zhipu is on the right path, but it needs to develop unique capabilities that set it apart from its competitors. Otherwise, in this highly competitive AI market, it will be constantly playing catch-up. The analysis is written in clear language, making it easy for non-financial professionals to understand.