China AI Unicorns MiniMax and Zhipu AI File for Hong Kong IPOs, Diverging on Strategy
Two leading Chinese generative AI startups have filed for initial public offerings in Hong Kong within a 48-hour window, presenting global investors with distinct investment theses: a consumer-focused global application strategy versus a domestic enterprise-grade infrastructure play.
MiniMax and Zhipu AI submitted their prospectuses on December 21 and December 19, respectively. The filings offer the first granular look into the financials of China’s rapidly evolving AI sector, revealing aggressive spending to chase state-of-the-art capabilities, rapid revenue scaling, and heavy backing from technology titans including Alibaba and Tencent.
The disclosures highlight contrasting paths to monetization in a capital-intensive industry. MiniMax reported a gross margin expansion to 23% in the first nine months of 2025, driven by subscription growth in its consumer apps, while Zhipu AI maintained a 50% gross margin through localized enterprise deployment, despite suffering from intense price wars in its cloud-based services.
These listings mark a pivotal shift from technological competition to commercial verification for China’s AI "unicorns." Investors must now weigh MiniMax’s ability to sustain user engagement against the steep compute costs of consumer traffic, versus Zhipu AI’s capacity to build a defensive moat in the enterprise market while offsetting massive infrastructure R&D expenses.
Consumer Super-Apps vs. Enterprise Infrastructure
The prospectuses reveal fundamentally different business models. MiniMax positions itself as an "AI Native Product" company. Its revenue from AI native applications surged to US$38.02 million in the first nine months of 2025, accounting for 71.1% of its total revenue, up from US$15.6 million in 2023.
In contrast, Zhipu AI operates as a technology enabler focused on the business - to - business (ToB) market. As of June 30, 2025, its revenue from localized (private) deployment reached RMB 162 million (US$22.3 million), accounting for 84.8% of its total revenue, up from RMB 26.3 million in the prior period, showing robust growth in the domestic enterprise sector.
Margin Dynamics and Market Efficiency
Gross margins serve as a key indicator of pricing power and technical efficiency for bot firms. MiniMax demonstrated a significant turnaround, swinging from a negative gross margin of 24.7% in 2023 to a positive 23.3% in the first nine months of 2025. Notably, its open platform business—which powers audio features for major digital reading and content platforms—recorded a gross margin of 69.4%. This suggests MiniMax has achieved efficiency advantages in model inference costs, likely utilizing its proprietary Mixture-of-Experts (MoE) architecture.
Zhipu AI maintained a high overall gross margin of roughly 50%, underpinned by the 59.1% margin of its localized deployment services. However, its cloud-based "Model-as-a-Service" (MaaS) business signaled distress. Margins for this segment plummeted from 76.1% in 2022 to negative 0.4% in the first half of 2025. The company attributed this decline to "lowering service prices in response to market trends," confirming that the price war in China’s API market has eroded profitability for standard cloud-based model services.
Aggressive R&D Spend
Both companies remain in a phase of heavy investment, though strategies differ in intensity. Zhipu AI is pursuing a "supersaturated" investment strategy. In the first half of 2025, the company reported R&D expenses of RMB 1.595 billion yuan (US$220 million)—more than eight times its revenue for the same period. The bulk of this capital was allocated to computing power service fees and personnel, reflecting a determination to close the gap with global state-of-the-art models.
MiniMax also maintains high spending but shows signs of operating leverage. Its R&D expense ratio narrowed to 337.4% in the first nine months of 2025 (US$180 million in spending against US$53 million in revenue), down from over 2000% in 2023. This indicates that as its consumer products scale, the company is beginning to dilute the immense fixed costs of model training.
Global Reach and Capital Backing
Geographically, MiniMax has established a global footprint, with the Chinese mainland contributing only 26.9% of its revenue in the first nine months of 2025. The United States market accounted for 20.4%, driven by the popularity of its overseas product, Talkie. Zhipu AI remains focused on the domestic market, serving the digitalization needs of China's government and enterprise sectors.
Both companies possess substantial capital reserves and powerful shareholder rosters. MiniMax held approximately US$1 billion in cash and financial assets, backed by Alibaba (13.66%). Zhipu AI had cash reserves of US$351 million, with a diverse cap table including Alibaba, Tencent, Meituan, Xiaomi, and state-backed entities like the National Social Security Fund.