Investing should rely on Golden麒麟 analyst research reports, which are authoritative, professional, timely, and comprehensive, helping you uncover potential thematic opportunities!

Carrying the title of the "First Large Model Stock," KNOWLEDGE ATLAS AI successfully debuted on the Hong Kong Stock Exchange. After a day of volatile trading, it closed up 13.17% at HK$131.5 per share, ending the session with a market capitalization exceeding HK$57 billion. The very next day, MiniMax closed up 109% at HK$345 per share, breaking the HK$100 billion market cap barrier and becoming the Hong Kong exchange's own "First Large Model Stock." In fact, KNOWLEDGE ATLAS also achieved a 20% gain on its second trading day, pushing its market cap to HK$69.8 billion, a fairly strong follow-up performance. However, some financial commentators attributed this to MiniMax's explosive surge igniting enthusiasm for the "large model" theme in the Hong Kong market. A thought-provoking question arises: With its strong backing from state-owned shareholders and a portfolio of orders from numerous local governments and leading enterprises, if KNOWLEDGE ATLAS had insisted on listing in the A-share market, could it have delivered a better performance? The answer might be quite different from most people's expectations. KNOWLEDGE ATLAS had previously attempted an A-share listing. On April 14, 2025, KNOWLEDGE ATLAS formally filed for tutoring with the Beijing Securities Regulatory Bureau, targeting the STAR Market, with China International Capital Corporation Limited (CICC) as its sponsor. According to the tutoring plan, KNOWLEDGE ATLAS was scheduled to complete all tutoring work by October 2025 and submit its listing application documents, aiming to file its IPO prospectus by the end of 2025 at the earliest. Subsequently, however, KNOWLEDGE ATLAS received no feedback, opinions, or inquiries from the China Securities Regulatory Commission (CSRC) regarding its proposed A-share listing. This "silent rejection" was a significant factor prompting the company to pivot directly to the Hong Kong market. Revisiting this question is not merely about comparing the A-share and Hong Kong markets; it's about considering what truly constitutes the optimal path for KNOWLEDGE ATLAS under the current regulatory environment, market preferences, and its own financial characteristics. Behind the silent rejection lie three major hurdles for unprofitable AI companies seeking to list on the STAR Market: First, the red line in scrutiny concerning the scale of losses and sustainable operational capability. Although the STAR Market's fifth set of listing standards permits unprofitable companies to list, it requires an "estimated market value of no less than RMB 4 billion" and that the "main business or products have been approved by relevant state departments, possess substantial market potential, and have achieved阶段性成果 (stage results)." However, KNOWLEDGE ATLAS's financials show a pattern of escalating losses alongside growth: 2024 revenue was RMB 312 million, but the net loss soared to RMB 2.958 billion; in the first half of 2025, revenue was RMB 191 million, yet the loss reached RMB 2.358 billion, nearly matching the full-year 2024 figure. Furthermore, as of June 2025, KNOWLEDGE ATLAS held only RMB 2.55 billion in cash on its books, while the first half alone saw an operating cash outflow of RMB 1.33 billion. This easily raises doubts about its ability to continue as a going concern under A-share listing scrutiny. Scrutiny on the STAR Market consistently emphasizes the controllability of losses. KNOWLEDGE ATLAS's R&D expense ratio skyrocketed from 147% in 2022 to 835% in the first half of 2025. This burn-rate-for-market-share approach is undoubtedly viewed as high-risk in the context of A-share listing reviews. Second, the R&D expenditure dilemma: computing power costs cannot be capitalized. This is a key technical obstacle. Of KNOWLEDGE ATLAS's R&D expenditure of RMB 1.59 billion in the first half of 2025, computing power service fees accounted for a hefty 71.8% (RMB 1.145 billion). A brokerage analyst indicated that under A-share accounting standards, computing power leasing is classified as an expense and cannot be capitalized as an intangible asset, unlike clinical trial investments for biopharmaceutical companies. This results in KNOWLEDGE ATLAS's massive R&D spending directly reducing its profits, making its balance sheet appear increasingly precarious. The higher flexibility in calculating R&D expenditures under Hong Kong's Chapter 18C rules was also a significant factor in KNOWLEDGE ATLAS's choice. Third, the fatal mismatch between the review cycle and time cost. The STAR Market exercises extreme caution when reviewing unprofitable companies. At the end of August 2023, the CSRC announced a "phased tightening of the IPO pace," only formally announcing the "resumption of listing applications for unprofitable companies under the STAR Market's fifth set of standards" in June 2025. Even after the resumption, the average cycle from acceptance to a hearing before the listing committee still exceeds 18 months. KNOWLEDGE ATLAS's cash reserves were clearly insufficient to smoothly sustain a successful A-share listing process. In contrast, Hong Kong's Chapter 18C allows for confidential filing. KNOWLEDGE ATLAS went from passing its hearing on December 19, 2025, to listing on January 8, 2026 – a mere 20 days. This "lightning-speed listing" pace holds a fatal attraction for large model companies that are burning cash every minute. Even if KNOWLEDGE ATLAS had overcome all hurdles and successfully listed on the STAR Market, its market performance might still have lagged behind its Hong Kong debut. Firstly, companies listing under the STAR Market's fifth standard need to enter the "STAR Market Growth Sector," where the threshold for individual investors requires assets of RMB 500,000 and at least two years of investment experience, along with signing a separate risk disclosure statement. This effectively excludes the majority of retail investors. KNOWLEDGE ATLAS's B2B business model is highly concentrated and lacks a compelling consumer-facing narrative. This further shrinks the already narrowed pool of retail investors, making it difficult to expect good liquidity and nearly impossible to replicate the 1,164 times oversubscription achieved in Hong Kong. Secondly, investors on the STAR Market prefer "hard tech" sectors like semiconductors and medical devices. The valuation models for large language models have yet to gain widespread acceptance and validation there. If benchmarked against the average price-to-earnings ratios of semiconductor companies on the STAR Market, KNOWLEDGE ATLAS's loss-making status could have led to a significantly lower offering price, or even a failed issuance. Conversely, Hong Kong's Chapter 18C explicitly lists artificial intelligence as a supported sector, and international capital is already accustomed to valuing companies like OpenAI and Anthropic, which follow the "high-losses, high-growth" model. Beyond liquidity and investor preference, the Hong Kong listing offered KNOWLEDGE ATLAS two additional advantages. First, it strengthens the company's global expansion strategy. KNOWLEDGE ATLAS's overseas revenue grew from 0.5% in 2024 to 11.1% in the first half of 2025 (primarily from Southeast Asia), and future growth requires reaching more international clients. As a bridgehead for international capital, the Hong Kong market can provide KNOWLEDGE ATLAS with a more international and diverse investor base, an advantage the STAR Market cannot readily offer. Second, the window of opportunity for large model companies to go public is rapidly closing. Competitors like MiniMax are all rushing towards IPOs. Had KNOWLEDGE ATLAS stubbornly persisted with the STAR Market, it risked missing out on the branding benefits of being the "first mover." In fact, the title of "World's First Large Model Stock" brought it immense attention, a label the A-share market could not have bestowed. KNOWLEDGE ATLAS's absence from the STAR Market is not a reflection of its lack of merit, but rather a fundamental misalignment between the A-share system's value preferences and the characteristics of AI large model companies, which require massive investment and have long gestation periods. The performance on its Hong Kong IPO's first and second days doesn't mean KNOWLEDGE ATLAS truly "lost"; it merely illustrates that differences in shareholder background and business narrative affect their reception in the Hong Kong market. Listing in Hong Kong was already the optimal solution for KNOWLEDGE ATLAS under the current circumstances.