Following the listing of "the world's first major large language model stock" KNOWLEDGE ATLAS (02513.HK) on the Hong Kong Stock Exchange on January 8, MiniMax (0100.HK) debuted on January 9, creating a back-to-back spectacle for two Chinese large model unicorns on the Hong Kong market, yet their stock price performances painted starkly contrasting pictures.
On January 8, KNOWLEDGE ATLAS listed at an issue price of HK$116.2 per share, opened at HK$120 per share for a 3.3% gain, temporarily fell below the issue price during the session, but rallied towards the close to finish at HK$131.5 per share, up 13.17% from the issue price, achieving a market capitalization of approximately HK$57.89 billion. Its Hong Kong public offering was oversubscribed by about 1,159 times, and its first-day trading could be described as "initially subdued then stabilizing," showing an overall rational trend.
A day later, MiniMax made its appearance on the Hong Kong Stock Exchange. With an issue price of HK$165 per share, its public offering was oversubscribed by 1,837 times and its international placement by over 36 times, indicating significantly higher demand than KNOWLEDGE ATLAS. It opened at HK$235.4 per share, surging 42.7% above the issue price, and closed with a gain of over 109%, corresponding to a market cap of HK$106.693 billion – making it the world's first large model company to exceed a valuation of HK$100 billion.
Meanwhile, KNOWLEDGE ATLAS, on its second trading day, performed much less impressively, closing at HK$158 with a market capitalization of HK$69.8 billion – a difference of roughly HK$37 billion between the two companies.
Why did the performance gap between these two leading domestic foundational large model companies, listing in Hong Kong around the same time and with similar IPO valuation benchmarks, turn out to be so substantial?
Institutional analysis suggests the divergence is closely linked to their respective business models, future monetization capabilities, and foreign investor appetite for allocating to Chinese AI companies. From a commercial strategy perspective, KNOWLEDGE ATLAS leans more towards providing a general-purpose AGI foundation and localized deployment solutions for government and enterprise clients, which typically entails a longer path to profitability. MiniMax, conversely, started with consumer-facing subscriptions and AI-native applications, presenting a more intuitive revenue growth trajectory that is arguably easier for the secondary market to "decipher."
Against the industry-wide consensus of high computing costs and persistent losses, Hong Kong investors have cast their financial votes, assigning different expectations to these two distinct large model commercialization paths.
An institutional investor in Hong Kong revealed that he, among others, had conducted valuation models and collectively believed KNOWLEDGE ATLAS's IPO valuation was set too high compared to MiniMax. Their models indicated a valuation gap of at least HK$150-200 billion in favor of MiniMax.
During the roadshow phase, both companies targeted valuations around HK$50 billion, but the disparity in the intensity of competition for allocation slots clearly indicated the perceived gap among institutional investors, which directly influenced their subsequent market performances.
Over 460 institutions participated in the MiniMax IPO subscription, resulting in over 70 times oversubscription (excluding the cornerstone portion).
This set a new historical record for institutional subscription in recent Hong Kong IPOs, surpassing the previous record held by CATL, which saw 30 times oversubscription (excluding cornerstones) upon its Hong Kong listing in 2025.
Regarding the fervent interest from foreign institutions in the MiniMax offering, some fund managers noted that while they recognize technical capabilities of leading Chinese large model firms like MiniMax still lag behind top US counterparts, the gap revealed in prospectuses was not as wide as initially imagined. More importantly, these foreign fund managers believed that compared to the sky-high valuations of US large model companies, MiniMax and similar Chinese firms were severely undervalued.
Huang Mingming, Founding Partner of Momentum Venture Capital, stated that MiniMax's listing attracted strong pursuit from global long-term funds, signifying market recognition of Chinese AI companies. He believes more severely undervalued Chinese tech firms will subsequently be discovered by a broader investor base globally. Momentum Capital has consecutively invested in MiniMax across six rounds since 2022, being the most frequent institutional participant.
Even including the cornerstone portion, MiniMax's oversubscription exceeded 36 times, far surpassing KNOWLEDGE ATLAS, which only saw over 15 times.
Some institutions and retail investors who failed to secure MiniMax IPO allotments turned to purchasing shares on the open market.
Even among retail investors, there was a significant difference in order placement for KNOWLEDGE ATLAS versus MiniMax. Public data shows 204,000 retail investors participated in the KNOWLEDGE ATLAS offering, whereas MiniMax attracted over 400,000.
A Hong Kong-based fund manager expressed little surprise at the lackluster institutional enthusiasm for KNOWLEDGE ATLAS. In his view, although both are general-purpose large models and currently in a cash-burning phase, investors are likely more focused on their future ability to generate and recoup capital.
Like most large model companies, both MiniMax and KNOWLEDGE ATLAS remain loss-making, which is not itself a primary concern for investors.
Financial data disclosed by KNOWLEDGE ATLAS shows an adjusted loss of approximately RMB 1.75 billion for the first half of 2025. Concurrently, KNOWLEDGE ATLAS held cash and cash equivalents of RMB 2.55 billion, and raised net proceeds of HK$4.17 billion from its IPO – implying a runway of about 1.9 years at the current burn rate without further public market fundraising.
For MiniMax, as of the end of September 2025, its loss was around US$180 million, while it held cash exceeding US$362 million.
Compared to the losses, investors are likely more attentive to the revenue and its composition for both KNOWLEDGE ATLAS and MiniMax.
KNOWLEDGE ATLAS's financial reports show revenues of RMB 57.4 million, RMB 124.5 million, and RMB 312.4 million for 2022, 2023, and 2024 respectively, representing a compound annual growth rate of 130%.
As of the end of June 2025, KNOWLEDGE ATLAS's clients included enterprises, public sector entities, and individuals, with institutional clients (enterprises and public sector) being the majority, exceeding 8,000 and covering approximately 80 million devices.
However, some investors also note that a significant portion of KNOWLEDGE ATLAS's revenue comes from the public sector, which could impact its future profit model. These public sector clients are predominantly government or state-owned enterprise-related entities. KNOWLEDGE ATLAS's reports indicate the vast majority of its revenue is sourced domestically, with localized deployment, and only a small portion of localized services cater to overseas clients, including those in Southeast Asia.
Within its localized deployment business, aside from the Internet and technology sector accounting for 38.3%, data from the first half of 2025 shows KNOWLEDGE ATLAS's public services segment contributed approximately 30%, second only to the internet/tech sector.
This revenue profile makes KNOWLEDGE ATLAS appear somewhat similar to the model of SenseTime from the previous AI wave – a government-focused business model, which raises concerns among some investors. SenseTime faced similar skepticism during its Hong Kong listing and subsequently delivered disappointing performance for a period, leading to losses for many investors.
Some Hong Kong investors previously indicated to Tencent Finance that MiniMax's business model is relatively clearer and is gradually achieving revenue diversification – boosting their confidence in MiniMax's future path to breakeven. MiniMax's prospectus shows monetization methods include subscriptions and payments for multiple products. Some investors even find the potential conversion rate of paying users promising, given MiniMax's existing 220 million users. They also see future overseas expansion as a promising avenue for MiniMax as a leading Chinese large model company.
A fund manager focused on AI industry investments noted that MiniMax's team is very young, which he considers another noteworthy aspect. Public data shows MiniMax has 385 employees with an average age of 29, and its founder, Yan Junjie, is only 37 years old.
In contrast, KNOWLEDGE ATLAS's co-founder and chairman, Liu Debing, is 49. More surprisingly, after the departure of original core founder Tang Jie, the role of Chief Scientist is filled by 90-year-old Tsinghua University Professor Zhang Bo.
Furthermore, the composition of their cornerstone investors differs significantly.
MiniMax's cornerstone investors are relatively more diverse, including not only mainland-backed institutions but also subscriptions from top foreign institutions like Mirae Asset, and sovereign wealth funds such as those from Abu Dhabi.
KNOWLEDGE ATLAS's cornerstone subscription portion also differs from most popular Hong Kong listings; it consists entirely of China-backed institutions. In recent years, hot southbound IPO projects, including Zijin Mining and CATL, almost invariably saw strong foreign investor participation.
While market enthusiasm for KNOWLEDGE ATLAS was not low, its published list of cornerstone allocations showed no foreign presence, whereas MiniMax, conducting its offering concurrently, featured several foreign cornerstone investors.
Additionally, KNOWLEDGE ATLAS's cornerstone subscription portion, unlike most popular Hong Kong listings, comprised solely of China-backed institutions.
Cornerstone investors accounted for 68.63% of KNOWLEDGE ATLAS's total offering, and although some were Hong Kong-based funds, they ultimately traced back to Chinese capital backgrounds.
It remains unclear how much of the allocation to the 221 institutions subscribing to KNOWLEDGE ATLAS was taken up by foreign investors. Tencent Finance learned that MiniMax's institutional orders included subscriptions from sovereign wealth funds in Europe, Singapore, the Middle East, Canada, and South Africa.
Beyond the companies themselves, this discrepancy might also relate to their sponsoring brokers. KNOWLEDGE ATLAS's sole sponsor was CICC, and all its underwriters were also China-backed.
MiniMax's intermediary institutions, besides CICC, were all top-tier foreign firms.
Naturally, these factors also influence IPO listing costs. Public data shows KNOWLEDGE ATLAS's listing expenses were approximately HK$174 million, while MiniMax's exceeded HK$221 million.
Comments