In-Depth Business Model Comparison and Investment Value Analysis of Zhipu AI and MiniMax

Based on comprehensively collected data, I will provide you with an in-depth business model comparison and investment value analysis report for Zhipu AI and MiniMax.

---

Zhipu AI was founded in 2019, spun off from the technological achievements of the Knowledge Engineering Group (KEG) under the Department of Computer Science and Technology of Tsinghua University, and is one of the earliest established general large model companies in China[1][2]. The company’s core team has strong research-oriented genes; CEO Zhang Peng joined the KEG Lab in 2002 and led the development of the scientific and technological intelligence platform AMiner[3]. Positioning itself as the “world’s first large model stock”, Zhipu AI officially listed on the Hong Kong Stock Exchange (HKEX) on January 8, 2026 (stock code: 02513.HK)[4].

MiniMax was founded in 2021, and its founder Yan Junjie served as Vice President of SenseTime; the core team also has a SenseTime background[3]. With the vision of “Intelligence with Everyone”, the company adopts a full-modal technical route and is one of only four technology companies in the world with self-developed full-modal large model capabilities[5]. The company is extremely young: as of the end of September 2025, its 385 employees have an average age of only 29, mainly consisting of post-1995 generation[6]. MiniMax listed on the Hong Kong Stock Exchange (HKEX) on January 9, 2026 (stock code: 00100.HK), becoming the world’s fastest large model enterprise to complete an IPO[5].

---

Dimension	Zhipu AI	MiniMax
Core Customers	B-end enterprise clients and G-end government agencies	C-end individual users and developers
Number of Customers	12,000+ enterprise clients	212 million individual users
Key Client Binding	9 of China’s top 10 internet companies are its users	Covers 200+ countries and regions
Market Focus	Mainly the Chinese market	Overseas markets account for over 70%

:

• Local Deployment
  : Contributed 84.5% of revenue in 2024, with a gross margin of 59%[7]
• Cloud MaaS Services
  : Accounted for 15.2% of revenue in the first half of 2025, with significant growth momentum[7]
• Revenue Scale
  : Revenues from 2022 to 2024 were RMB 57.4 million, RMB 125 million, and RMB 312 million respectively, with a compound annual growth rate (CAGR) of over 130%[2]
• First Half of 2025
  : Revenue reached RMB 191 million, surging 325.39% year-over-year (YoY)[2]

:

• AI-native Products
  : Contributed 71% of revenue in the first three quarters of 2025, mainly from Hailuo AI (USD 18.75 million) and Talkie/Xingye (USD 17.46 million)[6]
• Developer/Enterprise Services
  : Accounted for 28.9% of revenue, with a gross margin of 69.4%[6]
• Revenue Scale
  : USD 3.46 million in 2023 → USD 30.52 million in 2024 → USD 53.44 million in the first nine months of 2025[5]
• Growth Characteristics
  : 175% YoY growth in the first three quarters of 2025[5]

Company	Overall Gross Margin	C-end Business	B-end Business
Zhipu AI	50-56% (56.3% in 2024)	Approximately 50%	59% (local deployment)
MiniMax	23.3% (first three quarters of 2025)	4.7% (C-end)	69.4% (B-end)

From the gross margin comparison, it can be seen that MiniMax’s C-end business has extremely thin profit margins (its gross margin was -8.1% in 2024 and only turned positive to 4.7% in 2025), while its B-end business has stronger profitability[3]. Zhipu AI’s overall gross margin is more stable, reflecting the high-margin characteristics of B-end businesses.

---

:

• Adheres to a full-stack self-developed GLM (General Language Model) pre-training framework, using an autoregressive fill-in-the-blank design[7]
• Differentiated from GPT’s unidirectional attention architecture, it demonstrates advantages in long-text comprehension, logical reasoning, and low hallucination rates[7]
• Launched GLM-4.7 in December 2025, introducing an “interleaved thinking” mechanism, which performs outstandingly in programming and complex task planning[1]

:

• MaaS Platform
  : Over 2.7 million registered developers[1]
• Open-Source Models
  : Global downloads have exceeded 45 million[1]
• Agent Products
  : AutoGLM can independently operate smartphone and computer GUIs[1]
• Developer Tools
  : Zcode (AI code editor), Zread (code repository analysis tool)[1]

:

• Established a full-modal parallel technical route from its founding, and is one of the earliest teams in China to launch MoE (Mixture of Experts) models[6]
• Adopts a Mixture of Experts (MoE) structure, with significantly higher resource efficiency than peers[6]
• Cumulative spending since its founding is approximately USD 500 million, only 1% of OpenAI’s investment during the same period[5]

:

• Large Language Models
  : MiniMax M2 series, focusing on agents and code programming[6]
• Video Generation
  : Hailuo 2.3[5]
• Speech and Music Generation
  : Speech 2.6, Music 2.0[5]
• C-end Applications
  : Hailuo AI, Talkie/Xingye, MiniMax Voice[3]

---

Zhipu AI Revenue (in RMB 10,000):
2022: 5,740
2023: 12,500 (+118%)
2024: 31,200 (+150%)
2025H1: 19,100 (YoY +325%)

MiniMax Revenue (in USD 10,000):
2023: 346
2024: 3,052 (+782%)
2025Q1-Q3: 5,344 (YoY +175%)

:

• Net losses from 2022 to 2024 were RMB 144 million, RMB 788 million, and RMB 2.958 billion respectively[2]
• Net loss in the first half of 2025 reached RMB 2.358 billion, expanding 91% YoY[2]
• R&D expenditure in 2024 was RMB 2.195 billion, more than 7 times its revenue[2]
• Computing power service fees account for over 70% of R&D expenditure[2]

:

• Net losses from 2022 to 2024 were USD 73.7 million, USD 269 million, and USD 465 million respectively[3]
• Net loss in the first three quarters of 2025 reached USD 512 million[3]
• Cumulative R&D expenditure is approximately USD 450 million (RMB 3.16 billion)[3]
• Cost control achieved remarkable results in 2025, with sales and marketing expenditure decreasing 26% YoY[3]

:

• High R&D investment has put pressure on cash flow[2]
• Raised approximately HKD 4.35 billion in its IPO[8]

:

• Cumulative spending from its founding to September 2025 is USD 500 million[5]
• Total financing from 7 rounds exceeds RMB 10 billion[6]
• Raised approximately HKD 5.54 billion in its IPO[5]
• 14 cornerstone investors, including Aspex, Eastspring, Mirae Asset, ADIA, and Alibaba[5]

---

Indicator	Zhipu AI	MiniMax
Offer Price	HKD 116.2	HKD 165 (upper limit)
Public Offering Subscription Multiple	1,159x	1,837x
International Placement Subscription Multiple	-	36.76x
Number of Subscribers	-	Approximately 417,600

MiniMax broke the 1,200x oversubscription record set by Kuaishou in 2021[5].

:

• Closed 13% higher on its first trading day, and briefly broke below the offer price intraday[8]
• Subsequently continued to surge: +20.61% on the second day, +31.4% on the third day[8]
• Skyrocketed 62% on January 12, 2026, with the stock price hitting HKD 258[9]
• Total market capitalization exceeded HKD 110 billion[9]

:

• Opened 42.7% higher on its first trading day, closing with a 109% surge[5]
• The stock price once exceeded HKD 300[10]
• Market capitalization exceeded HKD 100 billion[5]

Company	Market Capitalization (HKD)	Price-to-Sales Ratio (2026E)
Zhipu AI	Approximately HKD 91.7 billion	Approximately 30x PS
MiniMax	Approximately HKD 123.1 billion	Relatively high (driven by growth narrative)

---

:

• Local deployment business provides stable cash flow, with high customer unit price and strong customer stickiness[7]
• 9 of China’s top 10 internet companies are its users, with a solid market position[1]
• The data security needs of government and enterprise clients form a natural barrier[7]

:

• Full-stack self-development of GLM, with original underlying architecture[7]
• Improved open-source ecosystem with a solid developer foundation[1]
• Leading layout in the agent field[1]

:

• 30x PS is lower than that of some peer AI companies[1]
• Revenue structure transformation (from local deployment to cloud) is expected to unlock growth potential[7]

:

• Losses continue to expand, with a loss of RMB 2.358 billion in the first half of 2025[2]
• High proportion of computing power costs (over 70%), greatly affected by chip supply[2]
• Gross margin of cloud business is under pressure[7]

:

• 73% of revenue comes from overseas markets, with less impact from domestic competition[8]
• 212 million individual user base, leading in user scale[5]
• Overseas users have stronger willingness to pay, with great room for customer unit price increase[3]

:

• Paid users surged 15 times in two years[6]
• Average spending per paid user of Hailuo AI increased from USD 36 in 2024 to USD 56 in September 2025[3]
• Product matrix covers full scenarios of text, voice, video, and virtual social interaction[5]

:

• Revenue growth is rapid, with 782% YoY growth in 2024[6]
• A trend of narrowing losses has emerged, with adjusted net losses tending to stabilize[3]
• The youngest management team, with better understanding of C-end user needs[6]

:

• C-end business has extremely low gross margin (only 4.7%), and the profit model has not yet been validated[3]
• User growth is highly dependent on product innovation capability, with high uncertainty[3]
• Global operations face multiple risks such as policy and compliance issues[5]

Dimension	Zhipu AI	MiniMax
Growth Certainty	★★★★☆	★★★★☆
Profitability	★★★☆☆	★★☆☆☆
Technical Barriers	★★★★★	★★★★☆
Market Space	★★★★☆	★★★★★
Valuation Reasonableness	★★★★☆	★★★☆☆
Risk-Adjusted Return	★★★★☆	★★★★☆

---

1. The business models of the two companies are inherently different
   : Zhipu AI follows a B-end deep cultivation route, while MiniMax follows a C-end globalization route, representing two viable paths for large model commercialization[3][8].

2. The capital market has given differentiated pricing
   : MiniMax has obtained a higher valuation premium due to its higher growth elasticity and globalization narrative, while Zhipu AI has received relatively rational pricing due to its stable B-end business[8].

3. Each of the two paths has its own advantages and disadvantages
   :
   • B-end Path (Zhipu AI)
     : Predictable growth, strong customer stickiness, stable cash flow, but slow expansion speed and greatly affected by government and enterprise budgets[7]
   • C-end Path (MiniMax)
     : High market ceiling, strong explosive power, low marginal cost, but profitability verification still takes time[3]

4. The industry is still in the early stage
   : The global large model market is expected to achieve a compound annual growth rate (CAGR) of 80.7% from 2024 to 2029, with broad market space[5].

: It is recommended to pay attention to

• B-end business provides a strong safety margin
• High technical barriers with a solid leading position
• Relatively reasonable valuation with favorable risk-return ratio

: May pay attention to

• Global layout has unique advantages
• Outstanding C-end product innovation capability
• Greater growth elasticity, but need to pay attention to profitability verification

: The losses of both companies are still expanding, the AI industry is highly competitive, and technology iteration is rapid. Investors need to closely monitor commercialization progress and changes in profitability.

---

[1] Wall Street CN - “First Domestic Large Model Stock” Zhipu Narrowly Rises on Debut: Model Iteration × Ecological Flywheel (https://wallstreetcn.com/articles/3762845)
[2] Securities Times - Zhipu Sprinting for “World’s First Large Model Stock”: Behind the AGI Dream (https://www.stcn.com/article/detail/3550275.html)
[3] OFweek AI Network - Zhipu and MiniMax Compete for “First Large Model Stock”: Each Has Its Own Challenges Amid High Growth (https://m.ofweek.com/ai/2025-12/ART-201712-8420-30677284.html)
[4] Securities Times Network - Two Bell-Ringings in 48 Hours: China’s AI Large Models Face Valuation Reassessment (https://caifuhao.eastmoney.com/news/20260114095446516258100)
[5] Guancha.cn - MiniMax’s Stock Price Doubles on Debut, Becoming the World’s First AI Listed Company with HKD 100 Billion Market Capitalization (https://www.guancha.cn/economy/2026_01_09_803325.shtml)
[6] The Paper - HKD 176.5 Billion: Capital Frenzy Over Two Large Model Giants (https://m.thepaper.cn/newsDetail_forward_32351566)
[7] China Business Journal - Zhipu vs. MiniMax HKEX Showdown May Ignite 2026 AI Listing Wave (http://www.cb.com.cn/index/show/zj/cv/cv135341401260)
[8] Jiemian News - Competing for “First Large Model Stock”: Zhipu Goes Left, MiniMax Goes Right (https://www.jiemian.com/article/13795470.html)
[9] Sina Hong Kong Stocks - Zhipu Skyrockets 62% in a Single Day, Total Market Capitalization Exceeds HKD 100 Billion! (https://finance.sina.com.cn/stock/hkstock/2026-01-12/doc-inhfzqar8417998.shtml)
[10] Securities Times Network - Surge Over 90%! Artificial Intelligence is Here! (https://www.stcn.com/article/detail/3583427.html)