A startup that bets on talking avatars and AI video just out-debuted nearly every Chinese tech listing of its generation. MiniMax (稀宇科技) hit the Hong Kong market on 9 January 2026 and saw its shares double on day one. Behind the pop is a cleaner story about who now bankrolls China's consumer-AI champions.
The listing that printed
MiniMax traded as 0100.HK. The tape:
- Offer price: HK$165; raised: about HK$4.8 billion (≈ US$619 million), the top of the range.
- Debut: shares doubled, closing at HK$345 and valuing the company at roughly US$13.7 billion — far above the ~US$6.5 billion IPO valuation.
- Demand: the Hong Kong public offer was oversubscribed about 1,209 times (some filings show ~1,837×); 14 cornerstone investors took roughly HK$2.72 billion (≈ US$349 million), about 70% of the deal.
- Cornerstone roster: Alibaba, Abu Dhabi Investment Authority (ADIA), Boyu Capital, Mirae Asset, Aspex and Eastspring, among others.
Founded in early 2022 by Yan Junjie (闫俊杰), a former SenseTime vice-president, MiniMax reached public markets just four years after inception — a pace the company itself called a record for the AI industry.
The private money, round by round
MiniMax's pre-IPO cap table reads like a map of China's strategic-tech investors:
- Series A (June 2023): led by Tencent at roughly US$250 million, valuing the firm around US$1.2 billion.
- Series B (March 2024): led by Alibaba at about US$600 million, lifting valuation to roughly US$2.5 billion. Gaming giant MiHoYo also invested.
- 2025 round: a US$300 million infusion led by Shanghai state-owned capital, pushing valuation past US$4 billion ahead of the listing.
- Total: more than US$1.15 billion raised before the IPO, across backers that also include Hillhouse, IDG, Redpoint-style VCs and Xiaomi-linked funds.
The arc is deliberate: consumer-internet giants provided the early scale capital, and Shanghai state capital provided the pre-IPO anchor — fitting for a company headquartered in Shanghai's Xuhui district, an AI cluster the local government actively cultivates.
Why investors paid up
Two things separated MiniMax from its "AI tiger" peers. First, a consumer footprint. Its apps — Hailuo AI for video generation and Talkie / 星野 for AI character interaction — reach global users; by September 2025 MiniMax reported over 212 million users across 200+ countries and more than 130,000 enterprise clients. Second, a multimodal model line treated as the product itself: the company's stance is "model is product," with a self-developed suite spanning text (MiniMax M2), video (Hailuo-2.3), speech (Speech-2.6) and music (Music 2.0), plus an open-source reasoning model, MiniMax-M1, notable for a 4-million-token context window.
Analysts framed the debut against Zhipu's a day earlier: Zhipu's enterprise-and-government model looked "stable but less exciting," while MiniMax's consumer growth "appealed more to investors seeking high-growth opportunities" (Omdia, cited by Reuters).
What the capital is for
About 90% of IPO proceeds are earmarked for R&D over five years — developing large models (大模型) and AI-native products. The rest covers working capital. The spending is necessary: MiniMax does not own compute and rents cloud capacity, and cloud costs drove a net loss of about US$512 million in the nine months to September 2025 on revenue of US$53.4 million. The bull case is that overseas subscription and enterprise API revenue (over 70% of sales, with overseas at ~73% in early data) compounds faster than the burn.
The honest caveats
MiniMax's story is not risk-free. The valuation doubled on debut to ~US$13.7B on a business with sub-US$100M trailing revenue — a multiple that presupposes global consumer AI adoption. Losses are structural, tied to rented compute rather than owned silicon. And the prospectus itself flags geopolitical and supply-chain exposure as a long-term constraint for any Chinese AI firm going global.
What readers can do now
- If you build consumer AI, study MiniMax's "model is product" discipline: it ships the model as the app (Hailuo, Talkie) rather than hiding it behind an API, a contrast worth benchmarking against your own GTM.
- If you invest or report, separate the debut pop from fundamentals — watch recurring overseas subscription revenue and cloud-cost ratio, not just user counts.
- If you use generative video/voice, test Hailuo AI and the open-weight M1 against Sora-class and open rivals; the multimodal breadth is the real differentiator.
Honest limitations
Core facts (HKEX listing 9 Jan 2026 as 0100.HK; offer HK$165; ~HK$4.8B raised ≈ US$619–620M; close HK$345, ~US$13.7B valuation; ~1,209× HK / ~1,837× retail oversubscription per filings; 14 cornerstone ~HK$2.72B incl. Alibaba, ADIA, Boyu, Mirae, Aspex, Eastspring; founded early 2022 by ex-SenseTime VP Yan Junjie; Tencent-led ~US$250M Series A June 2023 ~US$1.2B val; Alibaba-led ~US$600M Series B March 2024 ~US$2.5B val; ~US$300M Shanghai SOE 2025 >US$4B val; >US$1.15B pre-IPO; ~90% proceeds to R&D; 212M users / 130K enterprise / 200+ countries; 9mo-2025 revenue US$53.4M and net loss US$512M; Hailuo AI, Talkie/星野, MiniMax M1 4M context) come from Reuters (MarketScreener, ChinaNews, SCMP-sourced), Securities Times, 21st Century Business Herald, Dahe Cube and the company's prospectus. Pre-IPO round sizes and dates are consistent across Pandaily, NewsBreak and Chinese financial media; exact per-round figures should be read as reported, not audited. Currency uses ≈ 7.8 HKD/USD. This article is current to 29 September 2026 and is not investment advice.