For years, China's brightest AI labs watched their Western peers grab the IPO spotlight while they stayed private and quietly burned cash. This winter, the script flipped — and the venue was neither New York nor Shanghai, but Hong Kong. Within a single day, two of the country's most-watched generative-AI startups walked onto the exchange floor and into the public markets.
The message to global investors was blunt: you no longer have to buy the "AI concept stocks" when you can now own the labs themselves.
The first "large model" IPO
On 8 January 2026, Beijing-based Zhipu — formally Knowledge Atlas Technology JSC Ltd and known in China as 智谱 — listed on the Hong Kong Stock Exchange. The company bills itself as the world's first publicly traded firm built on general-purpose AI foundation models, a milestone widely described at home as the first "large model (大模型)" IPO.
The numbers were loud:
- HK$4.3 billion (≈ US$558 million) raised at an issue price of HK$116.2 per share.
- Shares opened at HK$120 and touched HK$130 intraday — about +11.88% — pushing market value above HK$57 billion.
- The Hong Kong public tranche was oversubscribed 1,159 times; the international tranche 15.28 times.
- 11 cornerstone investors — Beijing state capital, top insurers and major funds — took HK$2.98 billion, roughly 70% of the offer.
Zhipu was spun out of Tsinghua University in 2019 and counts Alibaba and Tencent among its backers. Chairman Liu Debing (刘德兵) framed the listing as the moment "general-purpose large models first reached the public market." The firm is still loss-making: it reported a net loss of about ¥2.36 billion (≈ US$330 million / HK$2.57 billion) in the first half of 2025, with R&D spend of ¥1.59 billion (≈ US$222 million) — roughly eight times its revenue for the period.
The 29-year-old upstart
One day later, on 9 January 2026, Shanghai's MiniMax Group (0100.HK) followed. It is the youngest of the cohort: founded at the end of 2021 by Yan Junjie (闫俊杰), a former SenseTime researcher. MiniMax is one of only a handful of global labs with a full multimodal model lineup in the top tier.
- Indicated raise of HK$3.8–4.2 billion (≈ US$490–540 million).
- 14 cornerstone investors — including Alibaba, Abu Dhabi's ADIA and Boyu Capital — subscribed HK$2.723 billion.
- The public tranche drew an eye-watering 1,209 times oversubscription.
- Per its prospectus, MiniMax had 212 million users across more than 200 countries, with overseas revenue above 70% of the total and revenue up 170% year-on-year in the first nine months of 2025.
Two of the so-called "AI six tigers (AI六小虎)" had gone public within 24 hours of each other — a first for China's model industry.
The chip and the brain
The wave was not only about model labs. Two other 2026 deals round out the picture of where China's AI capital actually flowed.
BiRen (壁仞科技, 6082.HK), a domestic GPU designer, listed on 2 January 2026 — the first GPU stock in Hong Kong and the largest "specialist technology" (18C) IPO there to date. It raised US$717 million before the over-allotment option, rising to US$825 million (≈ HK$6.4 billion) if fully exercised, at a market value north of HK$80 billion. Twenty-three cornerstone investors took about US$373 million.
And BrainCo (强脑科技), a non-invasive brain-computer interface (脑机接口) company founded by Harvard alumnus Han Bicheng (韩璧丞), closed a roughly ¥2.0 billion (≈ US$280 million / HK$2.18 billion) Series C in early 2026 — reported as the second-largest single brain-computer interface raise globally after Neuralink. Backers included IDG, Walden International, and contract manufacturers such as Lens Technology (蓝思科技).
Why Hong Kong, and what it signals
HKEX's own data tells the scale: across December 2025 and January 2026, 12 AI value-chain companies listed in Hong Kong, together raising roughly US$4.9 billion. They spanned infrastructure (chips), platforms (models) and applications — a far more complete ecosystem than the "concept stock" trades of previous cycles.
Hong Kong offered something the mainland exchanges and a US listing both struggled to provide at once: international capital access without the geopolitical scrutiny of a Washington float, plus a purpose-built 18C regime for pre-profit hard-tech firms. For labs burning billions on R&D, that combination is the point.
Honest limitations
Figures are drawn from HKEX, company prospectuses, and reporting by CGTN, China Daily, Shanghai Securities News, Xinhua and Hubei Daily. MiniMax's raise is the pre-pricing indicated range, not a final settled tally. BrainCo's round was reported by Chinese media; no independent audited filing was available to cross-check the exact amount, and the "second-largest after Neuralink" superlative comes from those reports. Intraday share moves (Zhipu's +11.88%, BiRen's first-day pop) are snapshots, not investment guidance. Most of these names are not yet profitable — Zhipu's ~¥2.36B H1-2025 loss and BiRen's pre-profit status show that a successful listing is not the same as a sustainable business. This article describes capital markets activity; it is not investment advice.
What readers can do now
- Track the HKEX new-economy pipeline. The remaining members of the "AI six tigers (AI六小虎)" and other model labs are the names to watch for follow-on floats.
- Separate the listing story from the earnings story. Before reading the hype, open the prospectus and look at R&D burn, path to profit, and overseas revenue mix.
- For the chip and brain-computer interface (脑机接口) tracks, watch policy and clinical or export milestones — not just financing headlines — since those are what will decide whether the capital translates into durable businesses.