Chinese Ai Company Ipo
The new year resolution for Chinese AI company started from Hong Kong IPO, Chinese AI giants MiniMax and Zhipu (Z.ai) recently launched their Initial Public Offerings (IPOs) on the Hong Kong Stock Exchange in early January 2026, with strong demand seen and trading set to begin next week.
- MiniMax, founded on 2021 by Junjie Yan, who was born in 1989, and an alumnus of Tsinghua University. Core Philosophy: Unlike some competitors who focus strictly on text (LLMs) or strictly on images, MiniMax positions itself as a company built on a “Multi-modal” foundation from day one (text, voice, and vision). Major Products and Models:
The “abab” Model Series: This is their family of Large Language Models (LLMs). They have released several iterations (e.g., abab 5.5, abab 6). They are known for having a very large context window (the amount of text the AI can remember at once) and strong performance in both English and Chinese. Voice/Speech (Audio-Gen): MiniMax is widely considered to have one of the best, if not the best, text-to-speech (TTS) engines in the industry. Their voice models are incredibly emotive, capable of sounding happy, sad, or angry, with very fast latency. This is a key differentiator for them. Hailuo AI (海螺AI): This is their main consumer-facing app (available internationally). It allows users to chat with their AI models, generate images, and, notably, create high-quality video clips from text prompts. Video Generation: In 2024, they released video-generation capabilities that rivals OpenAI’s Sora, allowing users to generate realistic video clips from text inputs.
Business Model:
To-C (Consumer): The Hailuo AI app. To-B (Business): They offer an OpenAI-style API platform, allowing developers to integrate their text, voice, and image models into third-party applications.
- Zhipu AI, founded in 2019, was born out of academia. It is a commercial spin-off from the KEG Lab (Knowledge Engineering Group) at Tsinghua University. Major Products: GLM Series: Flagship large language models including GLM-130B and the newer GLM-4 GLM-4V: Multimodal model capable of processing both text and images
Application Products: Zhipu Chat (智谱清言): Consumer-facing conversational AI assistant GLM Coding plan: AI-powered code generation and programming assistance tool. CogView: Text-to-image generation model CogVideoX: Text-to-video generation model
- Biren (The Semiconductor company): Founded by veterans from Nvidia on 2019, Biren already listed on the Hong Kong market on Jan 2nd. Major Products: Hardware: The “BR” Series of GPGPUs A. BR100 Series (The Flagship) The BR100 is Biren’s original, high-end product designed to compete directly with top-tier NVIDIA GPUs like the A100.Architecture: It uses a multi-chip module (MCM) design, stitching together two “chiplets” on a single substrate. This advanced packaging allows for massive performance.
B. BR104 Series (The “Sanction-Compliant” Version) Following US export restrictions in October 2022 that banned the sale of high-performance chips to China, Biren had to adapt. The BR104 is a modified version of the BR100 designed to comply with these new regulations.
Architecture: It is a single-chip design, derived from one of the chiplets used in the BR100. This naturally lowers its overall performance and inter-chip communication capabilities.
Software: The BIRENSUPA™ Ecosystem A chip is useless without software. Biren understands that NVIDIA’s dominance is as much due to its CUDA software platform as its hardware. Therefore, a massive part of Biren’s product is its software stack, BIRENSUPA.
So why go to Hong Kong? For a Chinese AI company today, the IPO decision is no longer purely financial; it’s a strategic one.
Listing in the U.S. is seen as too risky and potentially unsustainable. Listing in Mainland China provides domestic capital but lacks the international prestige, currency stability, and diverse investor base that a global tech leader desires.