---
title: "From KNOWLEDGE ATLAS to WENGE AI, Hong Kong stocks' AI welcomes the \"national team of large models\" from the Chinese Academy of Sciences"
type: "News"
locale: "en"
url: "https://longbridge.com/en/news/290996137.md"
description: "On June 26th, WENGE AI was listed on the Hong Kong Stock Exchange under Rule 18C, with a market value of approximately HKD 19.335 billion on its first day, being referred to as the \"first stock of decision-making large models.\" Its core narrative focuses on integrating large models into complex business scenarios to assist decision-making, marking a shift in the logic of the Hong Kong stock AI sector from basic technological capabilities to embedded business processes and commercial closed-loop capabilities"
datetime: "2026-06-26T09:41:03.000Z"
locales:
  - [zh-CN](https://longbridge.com/zh-CN/news/290996137.md)
  - [en](https://longbridge.com/en/news/290996137.md)
  - [zh-HK](https://longbridge.com/zh-HK/news/290996137.md)
---

# From KNOWLEDGE ATLAS to WENGE AI, Hong Kong stocks' AI welcomes the "national team of large models" from the Chinese Academy of Sciences

Author's Statement: This image is generated by AI
> "Core Tip"
> 
> What does the listing of the "first stock of decision-making large models" bring to the Hong Kong stock market?

**Author |** Zhou Ke

**Editor |** Liu Yang

On June 26, Zhongke Wenge (1956.HK) was listed on the Hong Kong Stock Exchange, with a market value reaching HKD 19.335 billion on its first day. This company, which completed its listing process through the 18C rule, has been labeled by the market with a tag that has never appeared in the Hong Kong stock market before—“the first stock of decision-making large models.”

In the past period, the Hong Kong stock market's understanding of AI companies has been more based on narratives around computing power infrastructure, model capabilities, and general technology platforms. Whether it is the supply of underlying resources or the output of model capabilities, the market's primary focus has often been on the technology trends themselves and the imaginative space brought by the expansion of the AI industry.

In contrast, Zhongke Wenge emphasizes how to integrate large models into complex business scenarios, participating in analysis, judgment, and decision-making processes. Therefore, the significance of Zhongke Wenge's listing is not only that the Hong Kong stock market has added an AI stock, but also that the AI sector in Hong Kong has for the first time seen a representative entity with "decision intelligence" as its core narrative.

Does this mean that, following the influx of AI companies to Hong Kong driven by the 18C rule, the measurement logic of the Hong Kong stock market for AI companies may be shifting from a focus on basic capabilities to placing greater importance on business process integration and commercial closed-loop capabilities?

**1\. 18C Fuels the AI Listing Wave in Hong Kong**

In March 2023, Chapter 18C of the Hong Kong Stock Exchange's "Listing Rules" was officially implemented, providing a listing channel for non-commercial specialized technology companies.

For most AI companies still in the high-investment stage, this is almost a naturally matching institutional design, where profitability is no longer a prerequisite, and technical attributes, R&D investment, and growth potential are the core thresholds.

Once the window opened, the Hong Kong stock market quickly became an important listing place for Chinese AI companies. Data released by the Hong Kong Stock Exchange shows that in the first quarter of this year, companies listed under 18C raised a total of approximately USD 2.5 billion, more than double the total amount raised in the past two years. By the end of May 2026, there will be 18 companies listed under 18C, the majority of which are AI companies. Major directions such as large language models, computing power infrastructure, AI chips, robotics, and autonomous driving have all seen representative companies listed on the Hong Kong stock market.

However, among this list, **the assets that have truly received high valuation pricing in the first round are still mainly concentrated in two directions: computing power infrastructure and general model platforms.** A direct evidence is the valuation. After the listing of KNOWLEDGE ATLAS, the PS (price-to-sales ratio) peaked at 594 to 733 times, and MiniMax's PS peak was also around 700 times Whether in terms of computing power or models, their commonality lies in being positioned at the forefront and having a compelling narrative, allowing investors to quickly incorporate them into the imagination framework of "next-generation foundational capabilities" without needing to deeply understand complex business details.

This also determines that the narrative of AI in the Hong Kong stock market during the first phase of 18C primarily answers the question of "Will AI become a sufficiently large industry?" rather than "What kind of value do different AI companies actually provide?" At this stage, the capital market is more willing to pay for technological prospects, industry space, and scarce positioning. The common pricing path for emerging industries entering the capital market is to prioritize infrastructure and general capabilities.

However, as the number of AI companies in the sector continues to increase, this coarse understanding will gradually become ineffective. Investors will no longer be satisfied with merely judging whether a company "is AI," but will further inquire, **What customers does it target? What scenarios does it embed in? What kind of business closed loop can it form?**

It is at this juncture that the "decision intelligence" represented by WENGE AI becomes particularly significant.

**2\. From "Can Generate" to "Can Decide"**

To understand the difference of WENGE AI, let's first look at a scenario.

On June 5, WENGE AI launched its general decision-making large model—Decitron Decision Machine. At the launch event, the decision machine addressed a question: "How might U.S. interest rate policy evolve after the change of the Federal Reserve Chair in 2026?"

If this question were handed to a general large model, the likely output would be a summary of the factors affecting interest rates. The approach of the Decitron Decision Machine is different: it further breaks down the question into "Under what conditions might it occur," "When might it occur," and "To what extent might it occur," deriving multiple paths, comparing different outcomes, and ultimately outputting a set of references for judgment.

This illustrates the difference between decision intelligence and general model capabilities. Computing power companies provide underlying resources, model companies output general capabilities, and tool products focus on efficiency improvement; whereas **WENGE AI's "general decision-making large model" focuses on analysis, judgment, and auxiliary decision-making after AI enters business processes.** It deals not just with "help me write a piece of copy" or "help me summarize a report," but with "After integrating multiple sources of information, does this risk need to be alerted?" "What are the various resource allocation plans, and what are their respective impacts?" "Under different changing conditions, how will the event evolve along which path?"

However, what is truly noteworthy is not just the launch itself, but WENGE AI's long-term accumulation in decision intelligence. For WENGE AI, the Decitron Decision Machine is a concentrated embodiment of the company's years of deep cultivation in the decision intelligence track, further productizing and platforming its technological accumulation, industry experience, and engineering capabilities.

The prospectus shows that **WENGE AI was founded in 2017 by a team of scientists from the Institute of Automation, Chinese Academy of Sciences, and has long been focused on enterprise-level large model-driven decision intelligence operating systems and services.** Since its establishment, the company has continuously delved into complex information analysis, cognitive intelligence, social computing, multi-agent systems, and AI-assisted decision-making, and has been consistently implemented in complex business scenarios such as public governance, finance, and industrial intelligence It is also in these scenarios that WENGE AI has gradually accumulated core capabilities such as data analysis, business ontology modeling, intelligent judgment, and multi-agent simulation.

These capabilities are organized into a complete technical architecture. The core foundation of WENGE AI is DIOS, which stands for Decision Intelligence Operating System. Around this system, the company has further formed a complete capability chain from the data middle platform X-Data, the industry large model Yayi, to the intelligent agent development platform DI-Brain. In other words, WENGE AI's core competitiveness lies in its systematic capability architecture formed around complex decision-making scenarios.

This is also where WENGE AI differs from general model companies. General models are better at providing open-ended understanding and generation capabilities, while decision intelligence often faces complex events, complex systems, and complex decisions: input information is highly multi-sourced, variables are interrelated, participants are not singular, and output results need to be interpretable, traceable, and able to truly enter the business judgment process.

Computing power and models are the foundation of AI capabilities, but what enterprises are truly willing to pay for continuously is often the specific value created after AI enters the business process. Decision intelligence precisely targets this layer.

**3\. What does "Decision Intelligence" bring to the Hong Kong stock market?**

As the number of AI companies under the 18C framework continues to increase, it has become difficult for the Hong Kong stock market to understand all targets with the single label of "Is it AI?" Companies with different capability levels and business models correspond to different valuation logics. The significance of WENGE AI lies in the fact that it adds a new category to the Hong Kong stock AI sector: decision intelligence.

What WENGE AI provides is the analysis, judgment, and decision-making assistance capabilities after models enter the business process. Therefore, what the market should really focus on is no longer just how strong the model is, but rather how deeply the client is embedded, how thick the scenario barriers are, whether the revenue quality is solid, and whether the productization path is clear.

**Globally, there is a relatively mature reference sample for decision intelligence, which is Palantir.** This company was founded in 2003, and its core business is embedding data analysis and software capabilities into the decision-making processes of governments and enterprises. When it went public in 2020, Palantir was still in a loss-making state, and its business model had strong characteristics of heavy redeployment and heavy delivery.

An important change occurred around 2023. With the launch of AIP (Artificial Intelligence Platform), Palantir's productization capabilities in the generative AI era were repriced, and the market's recognition of its "platform delivery + continuous order expansion" logic significantly increased. Subsequently, the company achieved GAAP profitability consecutively, with its profitability continuously improving, and its market value once exceeding $400 billion.

Palantir's trajectory illustrates two things: first, the path of decision intelligence is feasible, but it requires the accumulation of time; second, the key variable to success is not how strong the model is, but the speed of switching the business model—from customized projects to standardized products, reducing delivery costs and increasing gross margins **To some extent, the current position of WENGE AI is similar to the early stage of Palantir's transformation from heavy delivery to productization.** The company is in the early stages of transitioning from a "project-based" model to "productization." The delivery cycle has been compressed from 185 days to 80.2 days, and the gross margin has increased from 44% to 51.2%, indicating that this transition has begun. A net revenue retention rate of 139.5% shows that existing customers are continuing to increase their investments—this is the Chinese version of Palantir's "landing and expansion" model, first embedding the system into the core processes of clients and then gradually expanding orders.

There are several noteworthy figures in the prospectus. By the end of 2025, WENGE AI is expected to have served over 650 government and enterprise clients, including more than 190 Fortune Global 500 companies and their subsidiaries. Based on 2025 revenue, its market share among enterprise-level decision intelligence service providers driven by large models in China is 10.2%, ranking first. Revenue is projected to grow from CNY 249.7 million in 2023 to CNY 405.3 million in 2025, with a growth rate exceeding 60% over three years.

These figures indicate that WENGE AI is not just a pure concept; it has a certain commercial foundation. Compared to model companies that emphasize general capabilities, it has answered a key question: when large models enter real business processes, can they become a purchasable, deployable, and reusable decision support capability?

In this sense, WENGE AI's IPO is not just a financing event at the company level, but rather a signal that Hong Kong stock AI companies are beginning to be further differentiated. If 18C first addressed the question of "Can AI companies go public?", then **the next question the market really needs to answer is "How should different types of AI companies be understood, compared, and priced?"**

As the AI boom shifts from concept to realization, from narrative to operation, the market will increasingly value those companies that can reliably convert technical capabilities into business value and commercial results

### Related Stocks

- [01956.HK](https://longbridge.com/en/quote/01956.HK.md)
- [02513.HK](https://longbridge.com/en/quote/02513.HK.md)
- [00388.HK](https://longbridge.com/en/quote/00388.HK.md)
- [PLTR.US](https://longbridge.com/en/quote/PLTR.US.md)
- [80388.HK](https://longbridge.com/en/quote/80388.HK.md)
- [HKXCY.US](https://longbridge.com/en/quote/HKXCY.US.md)

## Related News & Research

- [China’s Zhipu AI sparks new ‘DeepSeek moment’ in Silicon Valley](https://longbridge.com/en/news/290925961.md)
- [China to match Anthropic’s top AI model in 2026, Zhipu founder tells Musk](https://longbridge.com/en/news/290463009.md)
- [An Interview with Barchart's AI Market Analyst CARL](https://longbridge.com/en/news/290709753.md)
- [1 of the Most Interesting AI Stocks in the Market Isn’t a Tech Company](https://longbridge.com/en/news/290606417.md)
- [Alibaba slashes Qwen AI model price during US working hours in fight for users](https://longbridge.com/en/news/290712852.md)