|
Neubauer Artists LLC Getting your Trinity Audio player ready... |
Chinese AI company MiniMax completed a HK$16 billion (approximately $2.0 billion) refinancing on July 10, consisting of a new share placement and zero-coupon convertible bonds, attracting over 20 international sovereign funds and long-term institutional investors with 7x oversubscription. The proceeds will primarily fund AI infrastructure and model R&D, alleviating financial pressure from cumulative losses of $2.68 billion. Founder and CEO Yan Junjie announced on the same day that he will forgo all compensation until MiniMax achieves AGI, while also committing 5% of his personal shares to incentivize the team and support the open-source community. The fundraising came one day after the company’s stock plunged 18% during its first major lock-up expiration, though core shareholders Alibaba and miHoYo stated they would not sell, while Goldman Sachs, Bank of America, and Citi issued buy ratings. The company is advancing development of its 2.7-trillion-parameter M3 Pro model and accelerating a shift from consumer to enterprise business, though long-term profitability remains unproven amid fierce industry competition and share price volatility.
Chinese artificial intelligence company MiniMax (0100.HK) has completed a large-scale refinancing amid dramatic share price swings. On July 10, the company announced it raised approximately HK$16 billion through a placement of new shares and issuance of zero-coupon convertible bonds, attracting participation from over 20 investors including international sovereign funds and long-term institutional investors.
The fundraising occurred one day after the company’s first major post-IPO lock-up expiration, during which the stock plunged nearly 18%. Market analysts view the move as both easing near-term funding pressure and signaling confidence in the company’s long-term value from core shareholders and management.
According to the announcement, the financing consists of two components: a placement of 35.6 million Class A shares at HK$268 per share, raising approximately HK$9.5 billion; and the issuance of HK$6.5 billion in zero-coupon convertible bonds maturing in July 2027. The transaction was 7x oversubscribed, with the initial offering size of approximately $1.8 billion ultimately expanded to over $2.0 billion driven by institutional demand.
The investor base spans Asia-Pacific, Europe, and US markets, including more than 20 long-term and sovereign funds. Multiple Pre-IPO and cornerstone investors also increased their positions in this transaction.
Use of Proceeds and Financial Pressure
MiniMax stated that 80% of the financing will be allocated to AI infrastructure and model R&D. From a financial perspective, this capital is critical for the company. For full-year 2025, total revenue was $79.038 million, while R&D expenses reached $253 million, with cumulative losses from 2022 to 2025 totaling approximately $2.68 billion. Prior to the IPO, the company held approximately $1.05 billion in cash and wealth management products on its books. At this burn rate, the HK$16 billion can sustain operations for roughly one to two years.
The financing will also result in equity dilution. The share placement and convertible bonds combined will add approximately 55 million new shares, representing about 17.5% of current total shares outstanding, posing potential pressure on the stock price going forward.
Founder’s Long-Term Commitment
On the same day as the financing announcement, MiniMax founder and CEO Yan Junjie issued an all-staff letter announcing a series of personal commitments.
“From today onward, until the day MiniMax achieves AGI, I will no longer receive any compensation from the company,” Yan wrote in the letter. Additionally, over the next four years, he will allocate shares equivalent to 4% of the company’s total equity from his personal holdings to reward team members who make long-term contributions, and commit another 1% of shares to establish a dedicated fund to continuously support open-source community development.
Yan Junjie was born in 1989 in Henan province and holds a PhD from the National Laboratory of Pattern Recognition at the Institute of Automation, Chinese Academy of Sciences. He previously served as a vice president at SenseTime. Co-founder Yun Yeyi also came from SenseTime and graduated from Johns Hopkins University. As of end-September 2025, the company’s 385 employees had an average age of just 29, with R&D personnel accounting for 73.8%.
Collective Shareholder Commitment and Stock Rollercoaster
Behind the financing announcement lies dramatic volatility in MiniMax’s stock price.
On July 9, the company faced its first major post-IPO lock-up expiration, with 153.5 million restricted shares flooding the market, representing nearly 49% of total shares outstanding at the time. Despite over 80% of Pre-IPO and cornerstone shareholders publicly stating their intention to hold long-term, and three major investment banks—Goldman Sachs, Bank of America, and Citi—simultaneously issuing “buy” ratings, the stock still tumbled nearly 18% on the expiration day, closing at HK$297.4, with approximately HK$20.5 billion in market value erased in a single day.
Market sources indicated that trading volume on the lock-up expiration day reached 20.944 million shares—roughly 6x the prior day’s volume—suggesting that selling by a small number of early shareholders was sufficient to disrupt the supply-demand balance in a market where the free float was previously under 6%.
To stabilize market confidence, two major early strategic shareholders, Alibaba and miHoYo, both stated they would “not reduce their holdings.” Alibaba expressed high regard for the MiniMax team’s technical capabilities and strategic resolve, and said it would continue deepening ecosystem collaboration in areas such as cloud computing and enterprise services. miHoYo emphasized that MiniMax is one of the very few companies globally to achieve cutting-edge capabilities in both language text and multimodal capabilities, aligning closely with its own long-term investment direction in artificial general intelligence. Additionally, the founding team has set a 12-month voluntary lock-up period, and over 80% of shareholders including Aspex, Boyu Capital, IDG Capital, and Janchor Partners have stated they will continue to hold.
Technology Roadmap and Commercial Transformation
Founded in late 2021, MiniMax has adhered to full-stack, in-house R&D across text, video, and voice modalities. Core products include the overseas AI companion application Talkie, AI video generation tool Hailuo AI, and an open platform providing API services to enterprise clients. As of end-2025, the company had cumulatively served over 236 million users across more than 200 countries and regions, as well as 214,000 enterprise customers and developers.
The company’s revenue mix is undergoing a significant shift. In 2024, consumer-facing revenue accounted for 71.4% and enterprise revenue 28.6%; by 2025, the consumer share declined to 67.2% while the enterprise share rose to 32.8%. Enterprise business grew nearly 198% year-over-year in 2025, significantly outpacing the 143% growth in consumer business. From a profitability perspective, in the first nine months of 2025, enterprise product gross margin reached 69.4%, far exceeding the 4.7% for consumer products. This indicates MiniMax is transitioning from an “AI application company” to an “AI platform company.”
On the technology front, the flagship M3 model was open-sourced on June 15, 2026, featuring 428 billion parameters (23 billion activated), utilizing a proprietary MSA sparse architecture, supporting native multimodality and million-token context windows, with inference speeds reaching 100 tokens per second. The company is developing M3 Pro with an expected parameter scale of 2.5 trillion to 2.7 trillion, planned for release and open-sourcing in the third quarter of 2026. If launched on schedule, this would become one of the largest open-weight models globally by parameter count.
Major Bank Endorsements and Industry Competition
Just one week before the financing announcement, Goldman Sachs, Bank of America, and Citi simultaneously issued “buy” ratings on MiniMax, with Goldman Sachs setting a HK$860 target price.
Goldman Sachs views DeepSeek V4’s upcoming introduction of differentiated peak-hour pricing as an early signal that the industry’s aggressive price war is moving toward rationalization. Against this backdrop, MiniMax’s M3 model, with a higher proportion of self-built optimized computing power and a more efficient architecture featuring smaller activated parameters, achieves significantly higher gross margins than peers. Goldman Sachs maintains its buy rating on MiniMax, citing that the M3 model sits in the ARR maximization quadrant of high token volume and attractive pricing, with current valuation at just 13x end-2026 estimated ARR, representing a clear discount compared to peer companies.
Bank of America noted the positive shift in MiniMax’s revenue mix from consumer to enterprise and cloud API businesses, and expects the company may be included in the Stock Connect program as early as August 6, bringing southbound capital liquidity support. Citi believes M3 ranks first among global open-source models on the Artificial Analysis Intelligence Index and GDPval-AA benchmark, and that the upcoming next-generation video model could serve as a key catalyst to reverse market sentiment.
In Goldman Sachs’ latest 50-page deep-dive report, the analyst team noted that Chinese AI large models are moving from “low cost” to “high intelligence.” The report describes China’s AI model market as forming a “two-tier structure”: the high-end market, represented by Zhipu GLM5.2 and Alibaba Qwen3.7 Max, is priced at approximately $1 per million tokens; the low-end market, targeting agent-task models, is priced as low as $0.06 to $0.2 per million tokens. Goldman Sachs forecasts that API and subscription revenue for Chinese AI models will grow from an estimated CNY 35 billion in 2026 to CNY 879 billion by 2030.
In terms of competitive positioning, Goldman Sachs identifies Zhipu and DeepSeek as the strongest in foundational text models, with ByteDance leading in multimodal capabilities. For MiniMax, its differentiated advantage lies in full-stack multimodal in-house R&D and global consumer-facing deployment, with 60% to 70% of revenue coming from overseas markets.
Risks and Challenges
Despite the successful fundraising, the challenges facing MiniMax have not dissipated.
On the stock price front, after briefly surging to a historical high of HK$1,330 in March, the shares had fallen over 70% by early July. Lock-up expiration pressure will persist through the next six months—another batch of approximately 56.96 million shares awaits release, representing about 18.2% of total shares outstanding, with lock-ups ending as late as around October 2026; by January 2027, shares held by the founder and core personnel will also gradually become unrestricted.
On the competitive front, MiniMax’s previously emphasized “value-for-money” positioning is losing effectiveness. Following DeepSeek V4’s release, the cost advantage has narrowed significantly. One user vividly commented: “Short tasks use GLM, long tasks use M3, planning use Claude,” highlighting MiniMax’s awkward position of not yet consistently reaching the top tier in model capability. Looking upward, it has not yet matched the capabilities of frontier models like OpenAI and Claude; looking downward, competitors like DeepSeek are driving prices even lower.
Additionally, 73% of the company’s 2025 revenue came from markets outside mainland China. According to media reports, relevant authorities are discussing the possibility of restricting overseas access to the most advanced domestic models. While the policy has yet to take shape, for a highly globalized model company, this has become a new valuation variable.
In the same week as MiniMax’s financing announcement, another Hong Kong-listed large model company, Zhipu, also launched an approximately HK$31.4 billion share placement on July 9. Two leading companies rolling out large-scale refinancing plans within just two days signals that the battle for resources in the large model industry is heating up. The HK$16 billion addresses MiniMax’s near-term funding gap, but over the long term, whether it can convert user traffic into sustainable payments, whether the enterprise business can truly scale, and whether it can translate value-for-money into revenue amid the dual squeeze of the parameter race and price war—these are the real points of focus for the market going forward.