Chinese artificial intelligence firm Zhipu made a highly anticipated debut on the Hong Kong Stock Exchange on Thursday, drawing overwhelming investor demand and setting a new benchmark for AI IPOs.
Shares closed up 13% at HK$131.50, giving the company a market capitalization of more than HK$55 billion ($7 billion), after its initial public offering was oversubscribed nearly 910 times.
The debut coincided with a similar surge of demand for MiniMax, another Chinese AI unicorn pursuing a Hong Kong listing. MiniMax closed its order window a day early due to overwhelming demand, with margin financing exceeding HK$253.3 billion and the public offering oversubscribed by 1,209 times.
“Being early has become the defining strategy for this generation of AI companies going public,” said Liu Zhou, founding partner and chairman of Fortune Venture Capital. “We invested in Zhipu as early as its Series A round at a valuation of 800 million RMB. By the IPO, that valuation had risen to 26 billion RMB. Early positioning secures a margin of safety and positions investors for significant upside.”
In an internal letter released alongside the IPO, Tang Jie highlighted Zhipu’s business performance and plans. The company’s cloud-based MaaS (Model-as-a-Service) platform, bolstered by the GLM-4.7 model, saw annualized revenue jump from RMB 20 million to over RMB 500 million in just ten months. The projected year-on-year growth rate for 2025 exceeds 900%, reflecting the rapid adoption of AI services in enterprise and government sectors.
Zhipu’s IPO raised HK$4.3 billion, with 70% earmarked for research and development and the remainder supporting platform expansion and marketing. Analysts said the move underscores the company’s ambition to scale while navigating the challenges inherent in commercializing AI technology still in its growth phase.
Zhipu’s IPO also rewarded early institutional investors. Legend Capital, which participated in several Series B rounds, holds a 6.73% stake. Based on the IPO issue price of HK$116.20, Legend Capital’s holdings are now worth HK$3.15 billion, a more than fourfold return.
Qiming Venture Partners, another early investor, participated in multiple B rounds, investing over RMB 150 million. Its 10.02 million shares are valued at HK$1.16 billion (RMB 1.04 billion), a near sevenfold gain. Meituan, which invested RMB 300 million in a Series B2 round, now holds 17.22 million shares worth HK$2 billion (RMB 1.79 billion), achieving a sixfold return. Other major corporate investors include Ant Group and Tencent, with stakes of 3.99% and 1.73%, respectively.
“The IPO not only provided a liquidity event for early investors but also injected capital to accelerate platform development and R&D,” said Wu Weijie, Zhipu’s vice president.
Zhipu’s revenue model relies heavily on enterprise (B-end) and government (G-end) clients. Cloud-based deployments, including API access, currently account for 15.2% of total revenue and operate at a gross margin of -0.4%, indicating early-stage losses. In contrast, private deployments for data-sensitive clients, which maintain local control over proprietary information, generate significantly higher margins. In the first half of 2025, Zhipu’s private deployment gross margin stood at 59.4%, slightly below historical levels of above 60%.
The company’s top five clients contributed 45.5% of total revenue in 2024, with the largest client accounting for 19%, highlighting the concentration risk common in B2B AI business models. Recognizable clients include Zhaopin, Kingsoft Office (WPS), and Mengniu Group, with nine of China’s top ten internet companies reportedly using Zhipu’s models.
Zhipu plans to increase revenue from lighter-weight API services and cloud-based offerings, aiming to raise their contribution to around 50% of total revenue. The strategy mirrors that of leading international AI companies, such as OpenAI and Anthropic, which monetize large models primarily through API subscriptions.
The AI Coding Plan, launched this year, generated an annualized recurring revenue of RMB 100 million, accounting for approximately half of Zhipu’s first-half revenue. While the service has been active for only two months, it signals the potential for growth in automated coding solutions amid rising demand for AI-assisted software development.
Zhipu is also pursuing international expansion. According to its prospectus, 10% of revenue from private deployments currently comes from Southeast Asia, while 90% remains domestic. Future growth is expected to include a larger proportion of overseas enterprise clients.
Additionally, the company may reduce reliance on G-end partnerships, which currently represent no more than 20% of total revenue. Social media posts from company executives suggest that Zhipu is prioritizing scalable, standardized MaaS deployments over labor-intensive, deeply customized solutions.
Zhipu’s IPO follows a wave of enthusiasm for AI stocks in Hong Kong and globally, reflecting the surging interest in generative AI technologies and large model applications. Investors are eager to back firms with strong R&D capabilities and early access to foundational AI models.
However, analysts caution that long-term profitability remains uncertain. The market for large AI models requires substantial capital to support ongoing model training, cloud infrastructure, and sales and marketing efforts. While the IPO provides Zhipu with funding to accelerate growth, the company faces the dual challenge of scaling rapidly while maintaining healthy margins.
“The ringing of the IPO bell is a milestone, but it also marks the beginning of a demanding journey for Zhipu,” said a Hong Kong-based analyst familiar with the company. “The firm must balance growth, profitability, and operational efficiency to sustain investor confidence and capture market share.”
Zhipu’s future trajectory will likely be shaped by its ability to expand cloud-based API services, maintain high-margin private deployments, and capitalize on overseas opportunities, all while navigating competition from domestic and international AI players such as MiniMax, Moonshot AI, and leading Western firms.
