Five months after a debut that saw shares double in a single session, MiniMax is posting accelerating revenue, approaching Hang Seng inclusion, and teasing a new flagship model that could reset its competitive position in China's crowded AI race.
When MiniMax listed on the Hong Kong Stock Exchange in January 2026, raising $619 million at a $6.5 billion valuation, the strategic choice of venue looked deliberate. US listings had grown complicated for Chinese tech firms, and Hong Kong offered both proximity to mainland capital and credibility with international institutional investors. What followed was a validation of that calculus that few observers could have scripted: shares closed up 109% on the first day of trading, peaking at HK$330 against an offer price of HK$165, pushing the company's market value briefly above $13 billion. Retail demand was almost absurd , the public tranche was oversubscribed 1,837 times.
The company behind that debut is still largely unknown to Western tech audiences. MiniMax was founded in early 2022 by Yan Junjie, a former SenseTime executive, alongside Yang Bin and Zhou Yucong. It builds multimodal AI models capable of processing text, audio, images, video, and music, and runs a suite of consumer products including Talkie, an AI companion app that ranked as the number one AI app by downloads globally by mid-2024, surpassing Character.AI with over 11 million monthly active users , more than half of them American. Its API business serves enterprises and developers, and that B2B segment was already generating 69.4% gross margins through the first nine months of 2025, even as the company ran net losses of $211 million during the same period.
The IPO attracted a serious cornerstone lineup. Alibaba, which had led a $600 million private financing round in early 2024 that valued MiniMax at $2.5 billion, returned as a cornerstone. Abu Dhabi's sovereign wealth fund ADIA joined alongside Aspex, Mirae Asset, and E Fund Management. The breadth of that roster , Chinese tech giants, Middle Eastern sovereign capital, and Asian institutional funds , was itself a signal of how MiniMax positioned itself: as a globally relevant AI infrastructure play, not merely a domestic product company.
The more interesting story is what has happened since January. MiniMax disclosed in late May that its annualized recurring revenue has more than doubled over just two months, crossing at least $300 million. Its enterprise client base now stands at over one million, a fivefold increase from six months ago. Global user numbers have reached 300 million. These are not metrics that typically follow a shaky IPO , they suggest the public market debut coincided with, and perhaps accelerated, genuine commercial momentum.
That momentum has earned MiniMax and Beijing-based Zhipu AI a place in Hong Kong's benchmark Hang Seng Index, effective June 5. Index inclusion matters for Chinese AI firms not just as a prestige marker but as a structural liquidity event , it forces passive funds tracking the index to hold the stock, providing a more stable shareholder base than the frenzied retail speculation that drove the debut pop.
The model side of the business is also evolving. MiniMax's current flagship, MiniMax-01, introduced a Lightning Attention mechanism with 456 billion parameters and a context window of up to 4 million tokens , twenty to thirty times longer than most competing models. The company has now teased its upcoming M3 series, built on a new sparse attention architecture that MiniMax claims delivers 15.6 times faster decoding speeds at one-million-token context lengths. That specific benchmark targets AI agent deployments, where processing very long inputs quickly and cheaply is a genuine commercial constraint rather than a benchmark-chasing exercise.
The competitive frame that matters
It is tempting to discuss MiniMax through the lens of DeepSeek, which reset the entire conversation about Chinese AI efficiency last year. DeepSeek released V4 in April and is reportedly exploring its first outside investment at a staggering $45 billion valuation. Moonshot AI, the company behind the Kimi assistant, is raising at north of $20 billion. Against those numbers, MiniMax at $13 billion on its best day looks modest.
But the comparison misses something. DeepSeek has stayed private and research-oriented. Moonshot AI is still pre-IPO. MiniMax, by going public, accepted transparency requirements and public market discipline that its rivals have not. Its financials are now on record. Its revenue trajectory is auditable. That accountability cuts both ways , it limits narrative flexibility , but it also signals a degree of commercial maturity that the others have not been tested against.
The Hong Kong choice also now looks prescient in a way the brief against US listings barely captures. Hong Kong gave MiniMax access to mainland Chinese institutional capital through the southbound Stock Connect, international institutional demand via the HKEX platform, and a regulatory environment that did not require untangling variable interest entity structures under US scrutiny. The 1,837-times retail oversubscription suggests the city's retail investor base was more than ready for a credible Chinese AI listing.
With M3 on the horizon, Hang Seng inclusion days away, and ARR compounding at a rate that could put the company past $400 million annualized by year-end, the question for investors is whether MiniMax can maintain margin quality as it scales compute to support its growing enterprise base. The 69.4% gross margin from late 2025 will face pressure as M3 requires more infrastructure investment. How that number moves over the next two quarterly reports will tell the real story of whether MiniMax is building a durable AI business or executing a well-timed capital markets play.
Also read: Samsung's last-minute bonus deal signals that AI's productivity windfall is becoming a labor negotiation rather than a corporate perk • The owner of 110 Pizza Huts is suing the chain, claiming 100 million in losses from the botch adaption of an AI tool • SentinelOne cuts 8 percent of its workforce and bets the proceeds on AI as investors punish the trade
