Analysis Report on Customer Retention Rate of Zhipu AI MaaS Platform

Based on the latest collected data, I will conduct an in-depth analysis of the customer retention rate of Zhipu AI’s MaaS platform.

According to Zhipu AI’s prospectus on the Hong Kong Stock Exchange and the latest financial data [0][1][2]:

Metric	2022	2023	2024	H1 2025
Revenue (RMB 100 million)	0.57	1.25	3.12	1.91
YoY Revenue Growth	-	119%	150%	325%
Net Loss (RMB 100 million)	1.44	7.88	29.58	23.58
Gross Margin	54.6%	64.6%	56.3%	50.0%
R&D Investment (RMB 100 million)	0.84	5.29	21.95	14.26

: In the first half of 2025, revenue reached RMB 191 million, with a year-on-year growth of 325%, demonstrating strong growth momentum. However, the net loss stood at RMB 2.358 billion, and the loss scale continued to expand [0][1].

---

According to verified multi-source data, the customer retention rate of Zhipu AI reaches

, significantly higher than the industry average of 65% [3][4][5]:

“Currently, 45 million global developers, 12,000 enterprise customers, and 80 million terminal devices form Zhipu’s large-scale large model ecosystem. On the OpenRouter platform, the API call volume of Zhipu exceeds the sum of all domestic models, and 9 of the top 10 domestic internet companies have become its customers.

The 78% customer retention rate far exceeds the industry average of 65%

”[5]

Metric	Data	Remarks
Number of Enterprise Customers	8,000+	As of H1 2025 [0]
Registered Developers	2.7 million+	As of H1 2025 [3]
Download Volume of Open-Source Models	45 million+	The most popular Chinese open-source model on Hugging Face [0]
Daily Token Consumption	4.2 trillion	Data as of November 2025 [3][4]
Revenue Contribution from Top 5 Customers	40%	Significantly decreased from 61.5% in 2023 [5]

---

• The GLM series models complete a base iteration every 3-6 months to maintain technological competitiveness [6]
• GLM-4.7 ranks among the top open-source models in international benchmark tests such as Code Arena [5]
• The model’s inference, programming, and Agent capabilities form differentiated competitive advantages [4]

• The open-source strategy effectively reduces customer acquisition costs, and the improved toolchain enhances user stickiness [3]
• Tools such as Zcode (AI code editor) and Zread (code library analysis tool) enhance developer retention [3]
• The commercial closed-loop of “open-source lead generation - tool-based retention - API monetization” is gradually operating [3]

• The revenue contribution from the top 5 customers decreased from 61.5% in 2023 to 40% in H1 2025, making the customer base more diversified [5]
• A trend of customers migrating from on-premises to the cloud has emerged, and the proportion of cloud revenue has increased from a low base in 2022 to 15.2% [3]
• The gross margin of customized services for government and enterprises reaches 72%, and high-margin businesses enhance customer stickiness [5]

• The application scale of China’s large model market is growing rapidly, reaching nearly RMB 50 billion in 2024 [7]
• Based on 2024 revenue, Zhipu ranks first among independent general-purpose large model developers in China, with a market share of 6.6% [0][3]

• Major players such as ByteDance (Doubao), Baidu (Ernie Bot), and Alibaba (Tongyi Qianwen) have launched fierce price wars [2]
• Relying on cloud service advantages and financial strength, major players have pushed API prices to the “floor level” [2]
• The gross margin decreased from 64.6% in 2023 to 50% in H1 2025, reflecting the impact of price wars [1][2]

• The top 5 customers still contribute 40% of revenue, posing certain customer concentration risks [0][1]
• With the expansion of revenue scale, trade and other receivables remain at a high level (RMB 453 million as of June 2025) [0]
• The prospectus clearly warns of “customer concentration and renewal risks” [3]

• The average monthly cash burn rate reaches as high as RMB 327 million [0]
• As of June 2025, the book cash and cash equivalents stand at RMB 2.55 billion, facing continuous funding pressure [0]

---

Scenario	Possibility	Key Driving Factors
Optimistic Scenario (Retention Rate > 80%)	Medium-Low	Continuous technological leadership, mature Agent ecosystem, breakthrough in overseas markets, eased price wars
Base Scenario (Retention Rate 75%-80%)	High	Maintain existing technological advantages, continuous operation of ecological flywheel, optimization of customer structure
Conservative Scenario (Retention Rate < 75%)	Medium	Intensified price wars, slowed technological iteration, loss of major customers

Based on the above analysis, I believe that the customer retention rate of Zhipu AI’s MaaS platform

, but it is subject to the following conditions:

:

1. ✅ Maintain the performance leading advantage of the GLM series models, with a continuous 3-6 month iteration cycle
2. ✅ Accelerate the commercialization of Agent products such as AutoGLM to enhance customer stickiness
3. ✅ Further reduce customer concentration, controlling the revenue contribution from top 5 customers below 35%
4. ✅ Increase the gross margin of cloud services (currently under pressure from price wars) to improve profitability
5. ⚠️ Address the impact of price wars from major players and maintain healthy customer acquisition costs

:

• 🔴 If price wars from major players continue, Zhipu may be forced to follow up with price cuts, compressing profit margins
• 🔴 If technological iteration slows down, customers may switch to competitors’ platforms
• 🟡 The effectiveness of overseas market expansion is uncertain

---

• Technological Barriers: Tsinghua University background, full-stack self-developed GLM architecture, continuous model iteration [3]
• Ecological Advantages: Open-source strategy + toolchain to build a developer moat [3][5]
• Market Position: Ranked first in revenue among independent general-purpose large model developers in China [0][3]

• Sustained Loss Risk: Recorded a loss of RMB 2.358 billion in H1 2025, and the profit path is not yet clear [0][1]
• Intensified Competition Risk: Price wars from internet giants compress industry profits [2]
• Customer Renewal Risk: The top 5 customers account for 40% of revenue, posing the risk of losing major customers [0]
• Cash Flow Risk: Monthly cash burn of RMB 327 million, continuous funding chain pressure [0]

With a customer retention rate of 78%, Zhipu AI has significantly exceeded the industry average of 65%, demonstrating strong customer stickiness. Supported by technological advantages, ecological flywheel effect, and customer structure optimization,

. However, achieving this goal requires continuous response to price war pressure, maintaining the pace of technological iteration, and accelerating commercialization capabilities. For investors, it is necessary to focus on the clarification process of its profit path and whether it can maintain a healthy growth curve in fierce competition.

---

[0] Guancha - “Zhipu AI Passes Hong Kong Stock Exchange Hearing: H1 Revenue Surges 325% to RMB 191 Million, Monthly Cash Burn Exceeds RMB 300 Million” (https://user.guancha.cn/main/content?id=1569130)

[1] Sina Finance - “Zhipu Submits Listing Application, H1 2025 Revenue Reaches RMB 191 Million” (https://finance.sina.com.cn/jjxw/2025-12-20/doc-inhcmhcr1525463.shtml)

[2] 199IT - “Zhipu AI: H1 2025 Revenue RMB 191 Million, Cumulative Loss RMB 6.2 Billion” (https://www.199it.com/archives/1803166.html)

[3] Wall Street CN - “‘First Domestic Large Model Stock’ Zhipu Narrowly Rises on Debut: Can Model Iteration × Ecological Flywheel Deliver Growth?” (https://wallstreetcn.com/articles/3762845)

[4] Sina Finance - “Soochow Securities: Zhipu (02513) From Tsinghua Lab to Hong Kong Stock AI Upstart, Focus on Model Iteration and Ecological Flywheel” (https://finance.sina.com.cn/stock/hkstock/hkgg/2026-01-08/doc-inhfqtyr6676639.shtml)

[5] Yicai Global - “‘World’s First Large Model Stock’ Goes to Zhipu, Telling the Capital Story of China’s Largest Independent Model Developer” (https://www.yicai.com/news/102981627.html)

[6] Sina Finance - “‘World’s First Large Model Stock’ Goes to Zhipu, Telling the Capital Story of China’s Largest Independent Model Developer” (https://finance.sina.com.cn/roll/2025-12-30/doc-inheqqfh7132794.shtml)

[7] Sina Finance - Industry Data Chart (https://n.sinaimg.cn/spider20251128/275/w1200h675/20251128/0dc8-7513f503abbca20dc7d3a88cbfd4bce3.jpg)