Chinese AI firms MiniMax, Zhipu reveal finances ahead of IPOs

Chinese AI firms MiniMax and Zhipu AI have shared their financial details for the first time as they prepare for Hong Kong IPOs.

The two startups, backed by major Chinese tech companies Alibaba and Tencent, aim to become the first local generative AI firms to go public.

Zhipu, which develops AI models and solutions for government-affiliated clients, reported 2024 revenue of 312.4 million yuan (US$44.4 million).

MiniMax, which offers global-facing AI products such as virtual companions and video editing tools, posted US$30.5 million in 2024 revenue.

Both MiniMax and Zhipu AI are valued at about US$4 billion each, much less than their US and European peers.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Zhipu AI made 312.4 million yuan (US$44.4 million) in 2024 from government-linked customers and developer tools 1. Its API (application programming interface) unit aims for half of revenue as it moves to standardized services 1. MiniMax booked US$30.5 million from global consumer apps such as virtual companions and video editors 1.
Both firms sit near US$4 billion in valuation with revenue about 300 to 400 times smaller than OpenAI’s estimated US$13 billion. Investors appear to bet on growth over current profits. MiniMax held USD 1.046 billion in cash in September 2025 2. Neither has shared gross margins, net losses, or compute costs (the cost to run AI models). Lack of these metrics hinders unit economics analysis.

Zhipu reports 2.7 million paying API users 3. GLM Coding Plan (a program to help developers build with its models) has topped RMB (Chinese yuan) 100 million in annual recurring revenue 3. SaaS companies and system integrators (IT consultancies that implement software) can join partnerships to embed Zhipu models in industry-specific apps for China-based clients. This fits China-based clients where government ties help.
MiniMax open sourced the MiniMax-M2 model, which ranked first on the Artificial Analysis benchmark, a public leaderboard for model performance 1. Developers can use its multimodal features for text, audio, video, and music 4 5.