On May 29, 2026, a notice was quietly posted on the official website of the China Securities Regulatory Commission: MiniMax has signed an A-share listing tutoring agreement with China Securities Co., Ltd. This comes just four months and twenty days after this AI unicorn debuted on the Hong Kong Stock Exchange. The transition from a Hong Kong listing to initiating A-share tutoring took MiniMax less than half a year. In the high-stakes sprint of the Chinese large language model race, no one wants to be left behind.
The Strategic Calculus Behind the A+H Move
Records on May 29, 2026, from the CSRC's IPO tutoring disclosure system show that MiniMax Group Inc. formally submitted its A-share IPO tutoring filing, with China Securities Co., Ltd. as its sponsor. Two days later, on May 31, MiniMax issued an announcement on the Hong Kong exchange, stating its board had resolved to explore preliminary proposals for issuing Renminbi-denominated shares and listing on the Shanghai Stock Exchange's STAR Market.
This marks the formal initiation of a dual-capital-market "A+H" layout for this general artificial intelligence company founded just over four years ago. Behind this lies a precise logic of capital relay. The Hong Kong market serves to establish a valuation benchmark, while the A-share market is intended to absorb subsequent assets; this order is crucial. Had the company applied for an A-share review concurrently with its January listing, the standard review process for an unprofitable AI company with no comparable peers or mature profit model on the STAR Market could have faced indefinite delays.
In fact, KNOWLEDGE ATLAS initiated this strategy even earlier than MiniMax. As early as April 2025, KNOWLEDGE ATLAS discreetly submitted its first A-share tutoring filing, three months ahead of the CSRC's July announcement that year regarding the targeted expansion of the STAR Market's fifth set of listing criteria. KNOWLEDGE ATLAS subsequently listed in Hong Kong, where its post-IPO share price surged over 13-fold, while MiniMax also achieved a 4-fold gain against the trend. Only after the Hong Kong market completed the global capital market's genuine pricing of large language model assets did these two companies turn to strengthen their positions in the A-share market—this is a precisely calibrated rhythm, not a spur-of-the-moment decision.
Unpacking the Triple Strategic Logic
Why pursue an A+H listing? Superficially, it's about "securing an additional financing channel," but a deeper analysis reveals the interweaving of three strategic calculations.
The first layer is dual-market financing to secure virtually unlimited "ammunition" for the capital-intensive nature of the business. The ultimate competition in the AI industry boils down to the speed of intelligence capability advancement, which is highly dependent on R&D efficiency. MiniMax's technology iteration cycle has been compressed to "monthly refreshes"—launching the M1 model in June 2025, the M2 and M2.1 models in the fourth quarter of the same year, and the M2.5 model in February 2026. The next-generation M3, Hailuo 3, and other full-modal models are already on the R&D agenda. Founder Yan Junjie candidly stated the company is simultaneously pursuing both model development and ecosystem building, noting, "Apart from major tech giants, we might be the only startup in Asia capable of doing both at once." Advancing on both fronts means the appetite for capital is insatiable.
Financially, MiniMax indeed urgently requires continuous capital infusion. Its total revenue for 2025 was $79.038 million, a year-on-year increase of 158.9%, but its adjusted net loss also reached $250.9 million, widening from the previous year. An A+H dual listing means the Hong Kong market focuses on absorbing US dollar and global institutional funds to support overseas business and global expansion, while the A-share STAR Market is favorable to hard-tech companies, helping them connect with domestic industrial capital and state-owned investors. The high sentiment and expectations of mainland investors for AI large language models have already pushed the overall valuation of the AI sector on the STAR Market above that of Hong Kong, with stronger liquidity. Utilizing this high-valuation window in the A-share market can provide MiniMax with a "dual-channel" capital guarantee.
The second layer is that the Hong Kong market has already provided a "valuation anchor," making the A-share listing less about financing and more about "realizing value." Market dynamics show that if a company's assets have been successfully priced in one public market, its listing in another becomes significantly simpler—due to the existence of mature comparable peers and a clear market reference point. This is precisely what A-share regulators need. MiniMax's Hong Kong IPO price was HK$165. On its debut, the share price surged approximately 109% to HK$345, subsequently rallying strongly to a peak of HK$1,330 on March 18, briefly becoming the highest-priced stock on the Hong Kong market. As of the close on May 29, its share price remained at HK$840, representing a cumulative increase of over 400% and a total market capitalization of HK$263.5 billion. This valuation anchor provides a strong reference point for A-share pricing and significantly reduces the risks associated with the A-share listing.
More importantly, in July 2025, the CSRC formally issued guidelines clarifying and expanding the scope of application for the STAR Market's fifth set of listing criteria, institutionally paving the way for AI enterprises to list on the STAR Market. The policy window is now open, and the first mover will gain a strategic advantage.
The third layer involves filling the gap in core A-share AI investment targets to lock in a "uniqueness" premium. Most AI listings on the A-share market are computing power companies, lacking core assets in general large language models. The simultaneous A-share push by MiniMax and KNOWLEDGE ATLAS will fill this gap in the core technology layer of AI large language models within the A-share market, shifting the valuation logic of the AI sector from "speculating on hardware" to "pricing core technology." This scarcity of such targets inherently implies a premium. Furthermore, with MiniMax's revenue primarily from overseas markets, accounting for 73% of its 2025 income, a STAR Market listing is expected to help address its shortcomings in expanding into diverse domestic scenarios and securing landmark projects.
Preparing for the Endgame: MiniMax's Unique Position and Industry Pressure
In the arena of domestic large language models, MiniMax holds a rare trump card—its "full-modal" technology roadmap. The company has independently developed a multi-modal model family covering text, speech, image, video, and music, including the M2 series text models, the Hailuo 2.3 video generation model, the Speech 2.6 voice model, and the Music 2.0 music model. This portfolio places MiniMax among the very few "all-rounder" large language model players globally.
Commercially, MiniMax has demonstrated an industry-rare growth pace. In 2025, revenue from native AI products reached $53.075 million, up 143.4% year-on-year; revenue from its open platform and enterprise services surged 197.8% to $25.963 million. In terms of user scale, the company's global enterprise and developer client base has surpassed one million, a fivefold increase from six months prior, with a global user base of approximately 300 million. The commercialization pace accelerated sharply in early 2026: over the past two months, the company's annualized recurring revenue doubled to at least $300 million, with the ARR doubling cycle compressed to just 60 days.
In terms of cost reduction and efficiency improvement, MiniMax has also shown remarkable resilience. Through algorithm optimization, operator implementation, and codec engineering iterations, the inference computing power cost per million tokens for its M2 series text models decreased by over 50% in February 2026 compared to December 2025. While R&D expenses increased by only 33.8% year-on-year, gross profit growth approached fivefold, indicating significantly improved R&D conversion efficiency.
Prior to listing, MiniMax completed seven rounds of financing, with its valuation soaring from $200 million to $4.24 billion. Top-tier institutions like Alibaba, Tencent, miHoYo, Hillhouse Capital, and Sequoia Capital made significant bets, providing strong endorsement for its listing. For the IPO, it secured 14 cornerstone investors, including Aspex, Eastspring Investments, Alibaba, and E Fund Management, with total subscriptions amounting to HK$2.723 billion.
Pre-listing financial data shows that as of September 2025, MiniMax had cumulatively spent $500 million on R&D and operations. However, concurrently, the company held approximately $1.1 billion in cash on its books. This means MiniMax possesses an "ammunition reserve" exceeding $1 billion, far surpassing its同期 losses. Its core motivation for listing is not merely "survival" but to accumulate energy for higher-dimensional competition.
However, substantial ammunition does not guarantee smooth sailing. The industry pressure facing MiniMax is rising exponentially.
Domestically, ByteDance has reportedly increased its 2026 AI infrastructure budget to 200 billion yuan, with its Doubao product launching paid subscriptions, firing the first shot in monetization. Giants like Baidu, Alibaba, and Tencent continue to ramp up investments, each forming complete product ecosystem closed loops. These giants possess longer time horizons and greater tolerance for error, allowing them to observe before going all-in—a capability difficult for independent large language model companies to replicate.
The independent large language model camp is also densely advancing capital actions: Moonshot AI completed a $2 billion financing round, with its valuation reportedly exceeding $20 billion, and is rumored to be preparing for a Hong Kong IPO; DeepSeek's first financing round reportedly values it at $45 billion; StepFun completed a nearly $2.5 billion financing round and is dismantling its VIE structure to sprint towards a Hong Kong IPO. The landscape of the昔日 "Hundred Models Battle" is consolidating towards the头部 at an astonishing pace. The race on this track has升级 to an上市冲刺, and no one wants to be left behind.
Internationally, OpenAI completed a $122 billion private financing round, with a post-money valuation of $852 billion and IPO preparations underway; Anthropic, after financing, reportedly surpassed OpenAI with a valuation of $965 billion. Global large language model companies are flooding into public markets at an unprecedented节奏.
Simultaneously, three major risks不容忽视. First, copyright litigation looms—US giants like Disney and Warner Bros. have sued "Hailuo AI" for infringement, alleging unauthorized use of film and television materials for model training. Second, a盈利拐点 is not yet visible; HSBC once warned that AI companies may need to持续烧钱 until 2030 before possibly achieving profitability. Third, the "dual-class share" structure—founder Yan Junjie holds Class B shares with 10 votes per share, controlling 59.98% of the company's voting rights. While this有利于 long-term decision-making, it also削弱了小股东制衡能力.
Among the previously much-discussed "Six Tigers" of large language models, apart from the already "上岸"的KNOWLEDGE ATLAS and MiniMax, those likely still at the base model poker table are只剩下 Moonshot AI and StepFun. The dream narrative of AGI is beginning to give way to clearer financial logic and capital return requirements. Industry competition logic is also rapidly转换—after two years of parameter races and price wars, "cost" is replacing "scale" as the new competitive focus, with investors开始更严格地 scrutinizing ARR growth rates, gross margins, and盈利节奏. The large language model赛道正从 "技术愿景"驱动转向 "财务指标"驱动.
For MiniMax, the A-share listing path is not without challenges. The company remains in a持续投入 stage, its A-share上市辅导 is still in its初期 phase, and a formal application will require navigating multiple审核环节. Additionally, the July expiry of the Hong Kong share lock-up period may bring阶段性抛压.
However, it is undeniable that once the dual-market platform is established, MiniMax will possess more ample ammunition, a longer runway, and more从容的节奏选择 in the endgame battle of large language models—and in this烧钱竞赛, having an extra筹码 often means an additional margin for survival.
From its inception in Shanghai in early 2022 to its Hong Kong listing in January 2026, from an IPO price of HK$165 to closing at HK$840, and then launching its sprint towards the A-share STAR Market just four months later—MiniMax has, in four years and five months,演绎了 one of the most激进 capital narratives in the history of Chinese AI entrepreneurship.
As the large language model赛道 enters a "winner-takes-all" endgame today, an A+H上市 is not merely a capital maneuver; it is a declaration of war向全行业宣告 "we are still at the table." However, Yan Junjie, holding a HK$263.5 billion market cap and over $2.6 billion in cash, perhaps understands better than anyone: in this endgame battle, valuation can be created overnight, but the true胜负手永远 lies within the code, the models, and the users.
Comments