Chinese artificial intelligence powerhouse Z.ai has filed for an initial public offering on the Hong Kong Stock Exchange, seeking to raise approximately $560 million. The company, formerly known as Zhipu AI, aims to become the world’s first publicly listed foundation model company, positioning itself as a significant contender against global AI leaders like OpenAI.
Z.ai’s Ambitious IPO Plans
Z.ai, officially Knowledge Atlas Technology Joint Stock Co. Ltd., is a prominent player in China’s AI landscape, recognized for its advanced large language models (LLMs) like the GLM series, which support both Chinese and English conversational interactions. The company also developed the Ying video generation model and AutoGLM, an agentic AI application for task automation on smartphones. Having previously secured substantial funding from tech giants such as Tencent and Alibaba, Z.ai’s IPO filing marks a significant step in its growth trajectory.
The company plans to issue 37,419,500 class H shares at an offer price of HK$116.20 (approximately $15.01), aiming for a total raise of $560 million and a valuation of $6.5 billion. If successful, this could represent one of Hong Kong’s largest stock listings. Z.ai intends to allocate 70% of the IPO proceeds to research and development, with the remainder designated for commercial offerings, partner ecosystem expansion, and corporate expenses.
Challenging Western AI Dominance
Z.ai’s latest model, GLM-4.7, is positioned as a direct competitor to offerings from Western AI firms. This 358-billion parameter model boasts GPT-4-level performance while significantly reducing costs, potentially by up to 80%. It features a sophisticated Mixture-of-Experts architecture and an extensive 200,000-token context window, enabling it to process large amounts of information without losing context. Notably, Z.ai has released the complete model weights for GLM-4.7 under an MIT license on Hugging Face, allowing developers to run it locally and eliminate ongoing API costs.
Performance benchmarks indicate GLM-4.7’s strength, with competitive scores in coding and math reasoning tasks against models like Claude Sonnet 4. The company’s aggressive pricing strategy, with its GLM Coding Plan starting at $3 monthly, offers a stark contrast to the pricing of many premium AI models. Z.ai anticipates that the bulk of its revenue will stem from enterprise customers, leveraging a model-as-a-service approach through its API platform, which currently serves over 2.7 million developers.
Financial Realities of AI Development
Despite its technological advancements and market ambitions, Z.ai’s IPO filing also sheds light on the substantial financial demands of developing advanced AI models. Similar to U.S. counterparts like OpenAI and Anthropic, Z.ai is experiencing significant cash burn. In the first half of 2024, the company reported an operating loss of $146.9 million on revenue of $6.4 million. This trend continued into the first half of the current year, with losses escalating to $271.1 million on revenue of $27.2 million. Research and development costs alone reached $227.5 million in the first six months of this year, underscoring the immense investment required to stay competitive in the AI field.
Navigating Geopolitical Headwinds
Z.ai has faced challenges, including being blacklisted by the U.S. Commerce Department in 2025 due to national security concerns. These sanctions restrict U.S. organizations from using its services and complicate its access to advanced processors. However, these measures do not appear to have significantly impeded the company’s progress, as evidenced by the continued development and release of its advanced models.
A Crowded IPO Landscape
Z.ai’s IPO is not the only one anticipated in the foundation model space. Reports indicate that rival company MiniMax is also preparing for an IPO, potentially listing next month. While MiniMax has not yet officially submitted its listing documents, its impending debut suggests a growing trend of AI companies seeking public market capital to fuel their expansion and innovation.