Analysis of Zhipu AI's Hong Kong Stock Listing Performance and Commercialization of Large Model Companies

Based on the collected information, I will provide an in-depth analysis of Zhipu AI’s Hong Kong Stock Listing performance and the commercialization challenges of large model companies.

---

Zhipu AI (Stock Code: 02513.HK) was officially listed on the main board of the Hong Kong Stock Exchange on

, becoming the world’s first listed company with general artificial intelligence (AGI) foundation models as its core business, hailed as the [1][2].

Indicator	Data
Offer Price	HK$116.20 per share
Opening Price	HK$120 per share (3.27% higher opening)
Closing Price	HK$131.5 per share (13.17% increase)
Total Market Capitalization	HK$57.89 billion
Over-Subscription Multiple	1159x
Net Proceeds	Approximately HK$4.173 billion

Looking at the first-day performance, Zhipu AI’s stock price experienced a

:

• Opening Session
  : Opened 3.27% higher at HK$120
• Intraday Correction
  : The gain narrowed at one point, and the stock price
  fell below the offer price temporarily
• Late-Session Surge
  : Closed with a 13.17% increase at HK$131.5[1][2]

This “rise first, then pull back, then surge” performance reflects the market’s complex attitude towards the company: while optimistic about the prospects of the AI track, it also has concerns about its commercialization capabilities and profit prospects.

Zhipu AI exhibits typical

characteristics[3]:

Year	Operating Revenue	YoY Growth Rate	Annual Loss
2022	RMB 57.4 million	—	RMB 144 million
2023	RMB 125 million	117%	RMB 788 million
2024	RMB 312 million	130%	RMB 2.958 billion
H1 2025	RMB 191 million	325%	RMB 2.358 billion

:

• Revenue maintains triple-digit growth, but
  losses are expanding at a faster pace
• Average monthly loss of approximately RMB 300 million in H1 2025
• As of H1 2025, cash and cash equivalents stood at RMB 2.552 billion, plus bank credit facilities totaling approximately RMB 8.943 billion[3]

---

Zhipu AI’s business model is mainly based on

, which specifically includes[2]:

• Local Privatized Deployment
  : Accounts for approximately 85% (project-based outsourcing)
• Cloud API Calls
  : Accounts for approximately 15%

:

• 70% (approximately HK$2.9 billion) for general AI large model R&D
• 10% (approximately HK$420 million) for optimizing the MaaS platform

Zhipu AI’s

, meaning every transaction is losing money[3]. This is mainly due to:

• Huge computing power investment required for large model training and inference
• The local deployment model requires a large amount of customized services
• Fierce market competition and price wars compress profit margins

• Purchase-Sales Inversion
  : The amount purchased from customers exceeds sales revenue
• Frequent Replacement of Major Customers
  : Most are one-time transactions with low stickiness
• Surge in Collection Cycle
  : Increased from 21 days to 112 days

• Squeeze from Internet Giants
  : Alibaba, Tencent, etc. are sweeping the market with free APIs and migration subsidies
• Technology Homogenization
  : The capability gap between foundation models is narrowing, with price becoming the main competitive dimension
• Accelerated Capital Consumption
  : Computing power service fees have soared 70x in three years, eating up 70% of R&D expenses

---

Based on industry practices, large model companies mainly have the following three commercialization paths[4]:

Path	Model	Representative Enterprises	Advantages	Challenges
API Service	Billed by token	OpenAI, Zhipu AI	Low marginal cost, rapid scaling	Fierce price wars, low profit margins
SaaS Subscription	Monthly/annual subscription	Claude	Stable revenue, high customer stickiness	Difficult to cultivate willingness to pay
Vertical Solutions	Customized services	AI companies in various industries	High customer unit price, clear demand	High labor costs, difficult to scale

Kai-Fu Lee, CEO of 01.AI, pointed out:

[5]. The core issues include:

“Business models that can drive revenue growth and profits are the new industry focus, and all large model companies need to be prepared.”

:

• Only
  2-3 pure foundation model companies
  may remain in the future
• Each company needs to find a
  differentiated positioning
  - vertical fields such as AI healthcare, AI finance, AI entertainment, etc.
• PMF (Product-Market Fit)
  has become a key verification indicator

• Focus on specific industries (such as legal, finance, healthcare)
• Establish domain-specific data and knowledge barriers
• Provide end-to-end solutions instead of pure model calls

• Shift from “selling models” to “selling capabilities”
• Develop autonomous task-executing agents
• Charge based on results (value sharing)

• Refer to DeepSeek’s
  545% cost profit margin
  practice[4]
• Optimize model inference efficiency and reduce marginal costs
• Open-source cost-reduction technologies and build ecological barriers

• Build an AI middle platform to improve the reuse efficiency of models and data
• Shift from project-based to platform-based intelligent capability output
• Build a developer ecosystem to achieve network effects

---

Zhipu AI has chosen the route of

:

• Localization of GLM Architecture
  : Compatible with over 40 domestic chips
• High-Frequency Iteration
  : Completes a foundation model iteration every 2-3 months
• International Layout
  : Gained paid access from 150,000 developers in 184 countries

Risk Factors	Opportunities
Cash flow pressure (average monthly loss of RMB 300 million)	HK$4.1 billion raised via Hong Kong Stock Exchange Chapter 18C listing
Negative gross profit margin	ARR (Annual Recurring Revenue) exceeds RMB 500 million (25x growth)
Fierce competition from giants	MaaS growth rate exceeds 900%
Business model to be verified	Brand effect of being the “world’s first large model stock”

Zhipu AI’s listing has

for the entire AI sector[2]:

• Verifies the IPO path for large model companies in the Chinese market
• Provides a reference for subsequent listings of AI companies (such as MiniMax)
• Undergoes commercial value testing on a more transparent capital stage

---

Zhipu AI’s first-day performance on the Hong Kong Stock Exchange (13.17% closing increase) indicates that the market remains enthusiastic about the AI track, but the volatile stock price reflects investors’

.

lies in:

1. Shift from technology-oriented to business value-oriented
   - Revenue that generates profits is sustainable
2. Find differentiated positioning
   - Avoid homogeneous competition and deepen vertical scenarios
3. Optimize cost structure
   - Improve inference efficiency and reduce marginal costs
4. Build customer stickiness
   - Shift from one-time projects to long-term service relationships

As Kai-Fu Lee said:

, but the premise is that every company can answer the core question of “whether there is a profitable business model”[5]. As the “world’s first large model stock”, Zhipu AI’s subsequent development will provide valuable reference experience for the Chinese and even global AI industries.

---

[1] Caixin - Zhipu AI Opens 3.27% Higher on Hong Kong Stock Exchange Debut, Up Nearly 9% with Market Cap of Approximately HK$55.5 Billion (https://companies.caixin.com/m/2026-01-08/102401577.html)
[2] Securities Times - “World’s First Large Model Stock” Zhipu AI Debuts with Market Cap Exceeding HK$57 Billion (https://www.stcn.com/article/detail/3580246.html)
[3] Eastmoney - Zhipu AI is Navigating the Most Dangerous Phase for Large Models (https://caifuhao.eastmoney.com/news/20251226100445247893150)
[4] OFweek AI Network - Three Commercialization Paths for AI Large Models (https://m.ofweek.com/ai/2025-03/ART-201700-8420-30658587.html)
[5] 36Kr - Kai-Fu Lee: 2025 is the Year of “Soul-Searching” for Large Model Companies’ Business Models (https://m.36kr.com/p/3116161197969416)