Analysis of Zhipu AI's Hong Kong IPO and Financial Status

Based on the information I have collected, I will provide you with a detailed analysis of Zhipu AI’s IPO plan to become the “World’s First Large Model Stock”, its financial status, and its commercialization challenges.

---

Zhipu AI officially passed the HKEX hearing on December 19, 2025, becoming the first company among the “Six Little Tigers of Large Models” to launch an IPO, striving to be the “First Large Model Stock” on the Hong Kong Stock Exchange [0]. According to the prospectus data, after the completion of the latest pre-IPO Series B6 financing round, the company’s post-investment valuation reached

[1].

Zhipu’s origins date back to 2019, founded by a team led by Professor Tang Jie from the Department of Computer Science at Tsinghua University, based on the technology transfer from Tsinghua University’s Knowledge Engineering Laboratory (KEG) [1].

Zhipu AI’s financial data shows the typical characteristics of an AI large model company:

.

[0][1]:

Period	Operating Revenue	Net Loss (Net Profit)	Adjusted Net Loss
2022	RMB 57 million	RMB 144 million	RMB 97 million
2023	RMB 125 million	RMB 788 million	RMB 621 million
2024	RMB 312 million	RMB 2.958 billion	RMB 2.466 billion
H1 2025	RMB 190 million	RMB 2.358 billion	RMB 1.752 billion
Cumulative over 3.5 years	Approx. RMB 684 million	Over RMB 6.2 billion	Approx. RMB 4.9 billion

Notably, the company’s average monthly loss in H1 2025 was close to

[1]. As of June 2025, the company’s cash and cash equivalents were only , and at the current loss rate, the cash reserves can only support operations for [1]. This is perhaps the core reason why Zhipu is eager to go public.

Despite severe losses, Zhipu’s commercialization progress is not slow:

• Compound Annual Growth Rate (CAGR) of Revenue
  : The CAGR from 2022 to 2024 exceeded
  130%
  [0]
• H1 2025
  : Revenue increased by
  325% year-on-year
  to RMB 190 million [0]
• Market Position
  : Based on revenue, Zhipu is the largest independent general-purpose large model developer in China, and the second-largest overall general-purpose large model developer in China, with a market share of
  6.6%
  [0]

The core source of losses is the

:

Year	R&D Expenses	Proportion of Computing Power Costs
2022	RMB 84 million	-
2023	RMB 529 million	-
2024	RMB 2.195 billion	70.7% (RMB 1.553 billion)
H1 2025	RMB 1.595 billion	Over 70%

Compared with international peers, Zhipu’s R&D expense-to-revenue ratio is as high as

, while this ratio for OpenAI in H1 this year is about , and for Anthropic [0]. The core reasons for the gap are:

1. Rigid growth in computing power costs
   : With the continuous expansion of large model parameter scales and the gradual increase in cloud users, computing power costs have become a rigid expenditure [0]
2. Lack of upstream strategic investment
   : OpenAI has strategic investment from Microsoft, and Anthropic has strategic investment from Amazon/Google, part of which is paid in the form of chips and computing power, greatly reducing the pressure of computing power costs [0]
3. Financing scale gap
   : Zhipu’s total estimated financing since its establishment exceeds RMB 16 billion, but there is still a large gap compared to the tens of billions of investment in OpenAI and Anthropic [0]

Regarding the “GSK partnership adjustment” you mentioned, my search found information indicating that Zhipu AI has a partnership with GSK in the field of AI drug research and development, but

[2].

From an industry perspective, GSK did make a number of strategic adjustments in 2025, including terminating certain clinical projects (such as the Phase III clinical trial of the TIGIT monoclonal antibody belrestotug), and in July 2025, reached a drug research and development cooperation with China’s Hengrui Medicine worth up to

[3]. This reflects the changes in strategic choices of pharmaceutical giants in the field of AI drug research and development.

Zhipu AI’s commercialization challenges can be summarized as follows:

• Industry analysis points out that “MaaS may be the worst business model in China in the short term; the more users, the more losses” [0]
• Cloud vendors have highly homogeneous business models and cannot reduce marginal costs through economies of scale
• Various vendors launch low-price/free strategies to seize market share, making it difficult to cover computing power costs

• The revenue share of the top five customers decreased from 55.4% in 2022 to 40% in H1 2025, but there is still a certain degree of dependence [1]
• The revenue share of the largest customer changed from 15.4% to 11%

• 16 rounds of financing have been completed in the primary market, with an estimated total of over RMB 16 billion, but capital is being consumed rapidly
• Going public has become a necessary option to obtain a more stable, low-cost financing channel
• However, the secondary market requires transparent financial status, and it is necessary to prove the feasibility of the business model to investors

• In H1 2025, COO Zhang Fan resigned, and the company fully integrated resources for government (G-end) and enterprise (B-end) clients
• The output of the toB team was significantly lower than that of the G-end team, leading to personnel optimization adjustments [0]

In its prospectus, Zhipu stated that it expects to reverse losses by increasing revenue and improving operational efficiency [1]. However, based on current data, achieving profitability requires breakthroughs in the following areas:

• Optimization of computing power costs
  : Reduce reliance on external computing power through self-developed chips or strategic cooperation
• Scaling of MaaS business
  : Expand API call revenue while ensuring profitability
• Opening up new revenue sources
  : Such as high-margin businesses like Agent products and enterprise-level solutions

---

Behind Zhipu AI’s pursuit of the “World’s First Large Model Stock” is a microcosm of the entire Chinese AI large model industry’s transformation from a “technology competition” to “commercial verification”. The RMB 6.2 billion loss over three and a half years reflects the industry’s widespread dilemma of

, especially the pressure of computing power costs. Adjustments to the GSK partnership (if true) may reflect the cautious attitude of international pharmaceutical giants towards AI drug research and development cooperation, which is another challenge for Zhipu, which relies on B2B commercialization.

Whether Zhipu can successfully go public and get out of the loss dilemma depends not only on its own technological iteration and commercialization capabilities, but also on whether the entire Chinese AI large model industry can find a sustainable profit path.

---

[0] 36Kr - “Losing RMB 300 million every month! From Zhipu’s prospectus, I see the cruel reality of large model competition” (https://m.36kr.com/p/3603323184743689)

[1] Phoenix Net Finance - “Losing RMB 6.2 billion in three and a half years, the RMB 24.4 billion large model giant targets Hong Kong Stock Exchange” (https://i.ifeng.com/c/8pIiJXX1M0K)

[2] Zhipu AI Prospectus and Public Disclosure Documents

[3] ByDrug Pharmcube - “In H1 2025, pharmaceutical giants abandoned 80 projects…” (https://bydrug.pharmcube.com/news/detail/0d7f85e7c2a5c913e4bb32b9d5381d94)