In-Depth Analysis of Zhipu AI (02513.HK) Hong Kong Listing

Based on the latest collected information, I will provide a systematic analysis of Zhipu AI’s Hong Kong listing debut performance, business model, and valuation for you.

---

Zhipu AI officially listed on the

, with stock code 02513.HK, becoming the world’s first public company focused on Artificial General Intelligence (AGI) foundation models as its core business [1][2]. The basic information of this IPO is as follows:

Indicator	Data
Issue Price	HK$116.2 per share
Total Fund Raised	Over HK$4.3 billion
Debut Closing Price	HK$131.5
Debut Gain	13.17%
Debut Market Capitalization	HK$57.89 billion
Hong Kong Public Offering Oversubscription Multiple	1159.46x
International Offering Oversubscription Multiple	15.28x

Notably, Zhipu AI briefly dipped below the issue price at the open, but quickly recovered and turned positive, with an intraday high of over 16% [1][3]. This trend reflects some market divergence on its valuation, but ultimately received positive recognition.

Zhipu AI’s cornerstone investor lineup can be described as an ‘all-star’ configuration, including [2][4]:

• State-owned institutions
  : State-owned assets from Beijing, Zhuhai, Chengdu and other regions
• Insurance capital
  : Leading insurance institutions
• Public funds
  : Large public fund institutions
• Industrial capital
  : Star enterprises such as Meituan, Ant Group, Alibaba, Tencent
• Well-known private equity firms
  : JSC International Investment Fund, JinYi Capital, Perseverance Asset Management, etc.

The 11 cornerstone investors subscribed for approximately HK$2.98 billion in total, accounting for nearly 70% of the total IPO fundraising amount, which indicates institutional investors’ recognition of Zhipu AI’s long-term value [4].

---

Zhipu AI adopts a

business model, providing general intelligence capabilities to developers and enterprises through API calls. This is significantly different from OpenAI’s revenue structure which mainly relies on consumer subscriptions [5][6].

:

Business Type	H1 2025 Revenue Share	Gross Profit Margin	Features
Localized Deployment	84.8%	59%	Serves government and enterprise clients, with data security guarantees
Cloud API Services	15.2%	Under Pressure	Strategic low-price strategy, strong marginal effects

as the core revenue source, mainly serving government and large-to-medium enterprise clients sensitive to data security. This business model has the advantage of generating stable cash flow, but its growth space is relatively limited [6].

are regarded by Zhipu AI as the key carrier of long-term value. In H1 2025, the daily token consumption reached 4.2 trillion. Soochow Securities predicts that the revenue share of cloud business will increase to 56% by 2027 [5]. Although the current strategic low-price strategy has put pressure on gross profit margins, the marginal effect is significant — as scale expands, unit costs will continue to decline.

Zhipu AI has built a unique

business closed-loop [5]:

High-performance open-source models → Attract global developers → Toolchain enhances stickiness → Convert to commercial orders

• Open-source strategy
  : As of H1 2025, global downloads of open-source models exceeded 45 million times
• Developer ecosystem
  : Over 2.7 million registered developers on the MaaS platform, and over 2.9 million enterprise and developer users on the API platform
• Toolchain layout
  : Launched applications such as Zcode (AI code editor) and Zread (code repository analysis tool) that deeply integrate the GLM-4.7 programming capabilities

This model effectively reduces customer acquisition costs while improving user retention and willingness to pay. Zhipu AI adheres to a parallel strategy of open-source and commercialization, establishing a differentiated competitive advantage globally.

Zhipu AI’s enterprise clients cover a wide range [5][6]:

• Key clients
  : 9 of the top 10 internet companies in China have become users of GLM models
• Institutional clients
  : Over 8,000 institutional clients with high data usage
• Market share
  : Based on 2024 revenue, Zhipu AI ranks first among independent general-purpose large model developers in China, with a market share of 6.6%

---

Zhipu AI exhibits typical characteristics of a high-growth tech stock — high growth accompanied by high losses [5][6]:

Financial Indicator	2022	2023	2024	H1 2025
Revenue (RMB 100 million)	-	-	-	1.91
YoY Revenue Growth	-	-	-	35.03%
Net Loss (RMB 100 million)	1.44	7.88	29.58	23.51

:

1. Expanding losses
   : Cumulative losses exceeded RMB 6.2 billion from 2022 to H1 2025, mainly due to a substantial increase in computing service fees and R&D expenses
2. R&D investment
   : Cumulative R&D investment during the same period was approximately RMB 4.4 billion, more than 8 times the cumulative revenue over the same period
3. Cost structure
   : In H1 2025, computing power costs accounted for 19% of total expenses, becoming the second largest expenditure after labor costs

According to Soochow Securities’ calculation, based on the IPO pricing, Zhipu AI’s expected

[5]. This valuation level has sparked market discussions:

:

1. Scarcity premium
   : As the ‘world’s first large model stock’, Zhipu AI is a pure large model target with irreplaceability
2. Technological leadership
   : GLM-4.7 ranks among the top in multiple international benchmark tests, and its paid API revenue exceeds the sum of all other domestic models
3. Growth space
   : The cloud business is in a period of explosive growth, with its revenue share expected to reach 56% by 2027
4. Institutional recognition
   : 11 cornerstone investors subscribed for nearly 70% of the shares, covering well-known institutions such as state-owned assets, insurance capital, and public funds

:

1. Performance losses
   : The net loss in H1 2025 was more than 12 times the revenue in the same period
2. Peer comparison
   : The 30x P/S ratio is higher than that of some peer AI companies, requiring rapid expansion of revenue scale to digest the valuation
3. Intensified competition
   : Surrounded by tech giants, technological leading advantages may be quickly eroded

According to the prospectus, the usage allocation of the approximately HK$4.3 billion IPO proceeds is as follows [6]:

Usage	Proportion	Description
General AI Large Model R&D	70%	Continue to invest in model iteration and computing power system construction
Optimize MaaS Platform	10%	Enhance cloud service capabilities
Develop Partner Network and Strategic Investment	10%	Expand ecological layout
Working Capital	10%	Daily operational needs

---

Against the backdrop of Zhipu AI’s listing, China’s large model industry is undergoing a drastic restructuring [6][7]. The once-famous ‘AI Six Tigers’ have clearly differentiated:

Company	Strategic Positioning	Current Status
Zhipu AI	Basic large models	Stick to pre-training track, listed in Hong Kong
01.AI	Basic large models	Withdrew from competition, pre-training team merged into Alibaba Tongyi
Baichuan Intelligent	Vertical fields	Shifted to medical AI
Moonshot AI	C-end applications	Focus on consumer-grade products such as Kimi
MiniMax	C-end applications	Listed in Hong Kong, over 80% gain on debut
JieYue XingChen	Dual-end layout	Covers both B-end and C-end

The industry consensus is that the players that can continue to participate in the basic large model competition may converge to the ‘Five Strong Foundation Model Players’ — DeepSeek, Alibaba, ByteDance, JieYue XingChen, and Zhipu AI [6].

Two tech giants, ByteDance and Alibaba, have fully entered the arena, bringing enormous pressure to independent large model vendors [6][7]:

Giant	Capital Expenditure Plan	AI Chip Budget	Strategic Focus
ByteDance	Approximately RMB 160 billion in 2026	Approximately RMB 85 billion	Full-stack AI layout
Alibaba	‘Three-year RMB 380 billion’ new infrastructure plan	-	Cloud computing + AI

In terms of talent competition, leading tech giants are sparing no expense. According to industry reports, ByteDance has become the ‘top payer’ for AI talents in China, and more than 10 core executives from the ‘Six Tigers’ system left to join tech giants in H1 2025 [6].

Zhipu AI’s technological advantages are mainly reflected in the following aspects [2][5]:

1. Self-developed GLM architecture
   : China’s first proprietary pre-trained large model framework, regarded as one of the few domestic architectures that can compete head-on with the GPT system
2. Model performance
   : GLM-4.7 tied with Anthropic and OpenAI models for first place in global coding in the Code Arena blind test
3. Multi-modal capabilities
   : Completed the layout of a full-stack model matrix covering language, vision, code, and intelligent agents
4. Computing power independence
   : Completed adaptation to over 40 domestic chips, with a first-mover advantage in terms of computing power independence and controllability
5. International layout
   : GLM-4.5 and GLM-4.6 have long ranked among the top 10 in global call volume on international platforms such as OpenRouter

However, the ‘half-generation gap’ in technological leadership is easy to close. In early 2025, DeepSeek shocked the market with a combination of open-source, free, and high-performance offerings, disrupting the original competitive rhythm [6].

---

Risk Type	Specific Performance
Profitability Pressure	Losses continue to expand, and revenue growth is unable to cover R&D and computing power investments
Valuation Digestion	A 30x P/S ratio requires rapid expansion of revenue scale to achieve reasonable returns
Price War	Fierce API price war in the industry, putting pressure on gross profit margins

1. Computing power constraints
   : Included in the US Entity List, facing difficulties in obtaining high-end computing power chips [6]
2. Talent loss
   : Tech giants poach talents with high salaries, raising concerns about the stability of the core team
3. Technological iteration
   : Model iteration every 3 months requires continuous high investment
4. Commercial conversion
   : There is no inevitable conversion chain between technological leadership and commercial revenue

1. Evolution of competitive landscape
   : The survival space of independent vendors under the squeeze of tech giants remains to be verified
2. Regulatory risks
   : Uncertainties in AI regulatory policies may affect business development
3. Technological route
   : Uncertainties exist in the technological route of large models, which may subvert existing advantages

---

:

• The MaaS model has strong scalability, and the marginal cost of cloud business continues to decline
• The developer ecosystem and institutional customer base lay a foundation for long-term development
• Localized deployment provides stable cash flow, reducing operational risks

:

• Current losses are severe, and revenue scale is far from sufficient to support the valuation
• Dependence on government and enterprise clients may lead to growth ceilings
• The marginal benefits of R&D investment are uncertain

: Zhipu AI’s business model has long-term logic, but it is difficult to achieve profitability in the short term. Investors need sufficient patience and risk tolerance.

Zhipu AI’s current valuation of approximately 30x P/S ratio is based on the following assumptions [5]:

1. The cloud business will maintain high-speed growth, with its revenue share reaching 56% in 2027
2. Technological leading advantages will continue to be converted into market share
3. R&D investment can maintain and expand competitive barriers

: If revenue growth falls short of expectations, price wars worsen due to intensified competition, or technological advantages are caught up, the current valuation will face significant adjustment pressure.

Dimension	Assessment
Short-Term	Debut performance met expectations, but be alert to the risk of valuation correction
Mid-Term	Focus on core indicators such as revenue growth rate, loss narrowing, and customer retention
Long-Term	Depends on technological breakthroughs, commercialization implementation, and evolution of industry competitive landscape

As the ‘world’s first large model stock’, Zhipu AI’s listing is a bellwether for China’s AI industry. Its price-to-earnings (P/E) and price-to-sales (P/S) ratios will become the ‘pricing anchor’ for primary market negotiations, guiding the industry narrative from ‘technological stories’ to ‘commercial value realization’ [5][6].

For investors with higher risk tolerance, Zhipu AI can be a configuration option in the AI large model track. However, given the high uncertainty of the industry and the current valuation level, it is recommended to remain prudent, control positions, and pay attention to subsequent technological iterations and commercialization progress.

---

[1] Guancha.cn - "First Large Model Stock" Zhipu AI Rises 13% on Debut, Company to Launch Next-Generation Model (https://www.guancha.cn/economy/2026_01_08_803199.shtml)

[2] Securities Times Network - Zhipu AI Lists: Unveiling the Capital Layout of the World’s First Large Model Stock (https://www.stcn.com/article/detail/3577855.html)

[3] Eastmoney - Zhipu AI and MiniMax Listings Ignite AI Track, Large Models Enter the New Stage of "Application is King" (https://caifuhao.eastmoney.com/news/20260109142902223461160)

[4] DeepTech - The World’s First Large Model Stock is Born! From Tsinghua Campus to Hong Kong Stock Exchange, Zhipu AI’s Six Years (https://www.mittrchina.com/news/detail/15747)

[5] Wall Street CN - "First Domestic Large Model Stock" Zhipu AI Narrowly Rises on Debut: Can Model Iteration × Ecological Flywheel Drive Growth? (https://wallstreetcn.com/articles/3762845)

[6] Tencent News - From 1159x Oversubscription to "Cash Crunch Breakthrough": Zhipu AI and Other Large Model Companies List, Firing the First Shot in the Survival Battle (https://view.inews.qq.com/a/20260108A02W0L00)

[7] 36Kr - MiniMax and Zhipu AI: Two Paths to Hundred-Billion IPOs (https://m.36kr.com/p/3636307697484805)

[8] Huxiu - Zhipu AI Becomes the World’s First Large Model Stock: How Foreigners See It (https://finance.sina.com.cn/stock/t/2026-01-12/doc-inhfyhif8915879.shtml)