The AI theme in Hong Kong stocks has been reignited. On June 15, KNOWLEDGE ATLAS (02513.HK) shares opened with a significant gap up and surged over 45% in early trading, reaching an intraday high around HK$1,620. By the midday close, the stock was still up approximately 28%, with trading volume notably higher, leading gains in the Hong Kong large model sector. At the market close, the Hang Seng Index rose 0.5% and the Hang Seng Tech Index gained 1.28%. Large model concept stock KNOWLEDGE ATLAS finished up over 32.5%.

The trigger for this rally is not entirely from the earnings season but is closer to a "model availability shock." As access to the most powerful overseas closed-source models faces tighter external controls, the market has begun shifting a premium for certainty towards domestic foundational model providers with controllable supply and localizable deployment. The most direct catalyst for this surge lies not in financial reports but in model availability itself.

Around June 12, U.S. AI company Anthropic was affected by an emergency export control order from the U.S. government, forcing it to temporarily suspend access to its latest and most capable models, Claude Fable 5 and Mythos 5, for all customers. Citing national security reasons, the U.S. implemented restrictions based on nationality, leading Anthropic to be unable to accurately verify the identities of global users in the short term, resulting in a complete shutdown of access to these two models, even affecting its foreign employees in the U.S. Anthropic expressed procedural objections to the decision and stated it is communicating with authorities to seek a more refined compliance mechanism to restore access.

For traders, the implication is straightforward: the most powerful overseas closed-source models are no longer just "expensive" but may become "unavailable" in certain scenarios or for certain entities. When enterprise clients and developers realize that "workflows could be cut off at any moment," capital quickly concentrates on domestic foundational models with controllable supply, localizable deployment, and comparable capabilities. KNOWLEDGE ATLAS has thus been pushed into the position of the "preferred domestic model stock."

In recent communications, institutions like Orient Securities used more direct language, stating that the delisting/restriction of Anthropic's leading models "will benefit leading domestic model companies by creating a substitution effect." A more critical catalyst comes from KNOWLEDGE ATLAS's own actions.

On June 13, KNOWLEDGE ATLAS announced the full public release of its most capable open-source large model to date, GLM-5.2, stating that APIs would be launched and the model officially open-sourced the following week, supporting a "truly usable 1M context" and continuing to emphasize leadership in long-context tasks. In a midday announcement on the Hong Kong exchange, the company confirmed the launch of GLM-5.2 and the 1M long context, clarified the MIT license open-source direction, and noted it would help increase usage on its open platform and API business. The company also emphasized it was "not aware of any undisclosed inside information."

At a moment when competitors might become "unavailable externally," KNOWLEDGE ATLAS has deployed a combination of moves—"a stronger model + lower migration barriers + a faster path to monetization via APIs"—leading the short-term market to re-evaluate it as an "option on rising usage volume."

Beyond fundamentals, today's high-volume surge is also supported by trading structure factors. The Shanghai Stock Exchange on June 5 and the Shenzhen Stock Exchange on June 8 successively announced that KNOWLEDGE ATLAS and five other stocks were added to the Southbound Stock Connect list, effective from June 8. Simultaneously, from June 8, KNOWLEDGE ATLAS was officially included as a constituent of the Hang Seng Tech Index.

This means systematic access for Southbound capital and passive allocation funds has been opened, while KNOWLEDGE ATLAS itself retains characteristics of a relatively limited free float and high stock price elasticity, making it prone to sharp fluctuations under thematic catalysts.

On another front, the market is trading on a "bad news is out" and "front-running" logic. Although facing share lock-up expiration pressure in July, capital broadly anticipates that as restricted shares become unlocked, KNOWLEDGE ATLAS's weight in relevant indices will be more accurately reflected, attracting passive fund allocation around the unlock period. This timing game between "selling pressure and buying power," combined with today's generally improved risk appetite, prompted short-term funds to gather rapidly.

Broader Implications for the Sector

Today's strong bullish candlestick is not an isolated thematic speculation but represents a market re-evaluation of a key trend: domestic large models are moving from policy advocacy towards commercial closure. At least three supporting pillars are simultaneously strengthening this shift.

From the supply side, as access to the most powerful overseas models becomes subject to export controls, domestic clients will prioritize "continuity" more highly. At this point, the biggest selling point for domestic foundational models is no longer just being "smarter," but being a reliable supply that "can run within the country, on local clusters, and under conditions where data does not leave the domain." Notably, on the hardware side supporting these models, the rise of domestic chips further strengthens the sustainability of domestic large models. Consequently, today's surge in KNOWLEDGE ATLAS also drove a broad rally in the Hong Kong chip sector.

From the demand side, the policy direction has shifted from "encouraging exploration" to "procurement implementation." From the Cyberspace Administration of China issuing guidelines for deploying government large models, to the State Council explicitly supporting the procurement of AI agent services, and the National Development and Reform Commission emphasizing the adaptation of domestic computing power, the institutional track has been laid.

In this context, KNOWLEDGE ATLAS, with the coding capabilities and long-context advantages of GLM-5.2, is well-positioned to capture the flow of "forced migration" traffic. Its transition from an "optional choice" to a "necessary choice" also signals that domestic large models are moving from being "demonstrable" towards a commercial closure of being "procurable and billable." Some analysts believe the domestic AI industry has formed a positive cycle of "model capability improvement—application expansion—token demand growth—sustained capital investment."

Challenges and Considerations Ahead

It is worth noting that a strong rally can easily make one overlook the harsh realities of valuation and cash flow. For KNOWLEDGE ATLAS, while today's "substitution narrative" is compelling, transitioning from sentiment re-evaluation to earnings re-evaluation requires crossing several concrete financial hurdles.

First is the validation of gross margins. The open-sourcing of GLM-5.2 and API price reduction strategy can indeed rapidly increase usage volume, but in the current environment of high computing power costs, whether economies of scale can cover expensive inference costs is key to determining the viability of the unit economics. Particularly for government and state-owned enterprise clients, customized deployment and private training often imply higher labor and operational expenses, requiring clear evidence of operating leverage improvement in subsequent financial reports.

Second is the pace of revenue recognition. From signing framework agreements to actually generating token consumption revenue, there is a long time lag involving proof-of-concept testing, security reviews, and budget approvals. The market is currently trading on "substitution potential," but investors ultimately value "revenue realization." As the July share unlock window approaches, during the process of share turnover, any rumors about slower-than-expected commercialization could be amplified.

Ultimately, the "new opportunity" for domestic large models cannot remain solely in the price-earnings ratio game of the secondary market; it must be grounded in a coherent and justifiable income statement. This will be the true test for KNOWLEDGE ATLAS in the next phase.