The hottest AI stock in the Hong Kong market recently is undoubtedly KNOWLEDGE ATLAS.

On June 22nd, KNOWLEDGE ATLAS shares continued their sharp ascent, surging nearly 40% intraday to reach a new all-time high. Within just a few trading sessions, the market has executed a dramatic re-rating of this company.

While the recent surge appears on the surface to be a burst of share price momentum, it is underpinned by three key drivers: stimulus from a new model, expectations for domestic substitution, and a revaluation of a scarce AI investment target.

Primary Catalyst: GLM-5.2

The key features of the new GLM-5.2 model are a 1 million token context window, enhanced coding capabilities, proficiency with long-range tasks, and open-source availability.

In essence, this is not merely a model that is better at casual conversation, but one more suited for complex work tasks. Examples include processing lengthy documents, large codebases, corporate knowledge repositories, code assistance, and agentic tasks.

This distinction is crucial. As large language models have evolved, the competition is no longer about who can write the best short essay, but about which models can genuinely integrate into workflows. The models that can help enterprises save manpower, boost efficiency, and handle complex tasks are the ones with true commercial value.

From this perspective, KNOWLEDGE ATLAS's positioning is relatively clear: its focus is not primarily on entertainment and consumer traffic but leans more towards being an enterprise-grade AI engineering foundation.

Research indicates this model has also garnered a positive reputation overseas.

Growing Expectations for Domestic Model Substitution

While powerful overseas closed-source models exist, their long-term use by domestic enterprises—especially in sectors like government, finance, energy, and manufacturing—raises concerns about data security, supply stability, and regulatory compliance.

In the past, domestic large models might have been merely a "backup option," but they are now evolving into a "strategic necessity." If enterprises require a stable, controllable, and locally deployable large model foundation in the future, companies like KNOWLEDGE ATLAS will be viewed in a new light by the market.

Therefore, this rally is not just about speculating on a new model release; it's about trading on a larger narrative: domestic AI infrastructure.

Rapid Revenue Growth

KNOWLEDGE ATLAS is not a concept company with no revenue. Previously disclosed figures show the company's revenue grew rapidly from 57.4 million yuan in 2022 to 124.5 million yuan in 2023, and further to 312.4 million yuan in 2024. Revenue for the first half of 2025 was approximately 191 million yuan, maintaining growth.

Its revenue primarily comes from large model services, including on-premises deployment and cloud-based deployment. On-premises deployment suits government and enterprise clients with higher contract values, while cloud-based APIs and MaaS platforms are lighter and more akin to the platform-based business model envisioned for future large model companies.

However, a significant issue is also apparent: while revenue growth is fast, losses are substantial.

The large model industry is inherently capital-intensive, requiring heavy investment in research and computing power. The stronger the model, the greater the need for continuous investment, and the larger the investment, the harder it is to avoid short-term losses. KNOWLEDGE ATLAS is not yet a profitable growth stock in the traditional sense but is an AI infrastructure company in a high-investment phase.

Consequently, the current surge in KNOWLEDGE ATLAS cannot be simply dismissed as pure speculation, nor can it be claimed that its financial performance has already been realized.

More accurately, the market is re-pricing KNOWLEDGE ATLAS from an ordinary AI company to a scarce domestic large model infrastructure investment target.

What truly needs monitoring going forward is not how much the stock price rises in a single day, but three key questions:

First, following GLM-5.2, can KNOWLEDGE ATLAS consistently maintain its position within the top tier of domestic model capabilities?

Second, can its model capabilities drive continued volume growth in API, cloud deployment, and enterprise client revenue?

Third, amidst high R&D and computing power investments, can the structure of its losses gradually improve?

If progress is made on these three fronts, KNOWLEDGE ATLAS could evolve from a short-term hot stock into a benchmark company within the domestic large model industry chain.

However, if subsequent developments are limited to model releases and market sentiment without tangible improvements in revenue, client acquisition, usage volume, and profitability, the stock could face significant volatility after a period of short-term overvaluation.