Within a 48-hour window, two Chinese AI unicorns submitted their IPO applications to the Hong Kong Stock Exchange. On December 19 and December 21, Beijing Zhipu Huazhang Technology Co., Ltd. (Zhipu AI) and Shanghai MiniMax Technology Co., Ltd. (MiniMax) disclosed their prospectuses, totaling nearly a thousand pages. These documents not only reveal the financial realities of China's AI large model companies but also highlight the stark contrast between their consumer-facing (ToC) and enterprise-focused (ToB) commercialization strategies.

**Business Model Debate: ToC vs. ToB** **MiniMax: C-End Driven, Betting on "Super Apps"** MiniMax operates more like an internet product company, branding itself as an "AI Native App" developer. Its flagship products include the multimodal interaction platform Talkie (overseas version) and Xingye (domestic version). According to its prospectus, AI Native Apps are MiniMax's core revenue driver. Revenue from this segment surged from $758,000 in 2023 to $21.8 million in 2024 and further to $38.02 million in the first nine months of 2025, accounting for 71.1% of total revenue.

This growth is fueled by explosive user expansion. Monthly active users (MAU) for Talkie/Xingye skyrocketed from 3.1 million in 2023 to 27.6 million by September 2025. More importantly, paying users reached 1.77 million in the first nine months of 2025, with average revenue per paying user (ARPPU) rising from $6 to $15. This demonstrates MiniMax's successful "user growth-retention-monetization" loop.

**User Data**: As of September 30, 2025, MiniMax's AI Native Apps have served over 200 million users, with MAUs hitting 27.64 million. **Monetization**: Primarily via subscriptions, in-app virtual purchases, and online marketing services. Talkie/Xingye leverages deep user interactions with virtual characters to foster strong engagement.

Notably, MiniMax's dominance extends beyond C-end apps—it’s also a hidden giant in B-end audio solutions. Its Speech-02 voice model is globally recognized as top-tier. In 2023, its largest client (contributing 37.2% of revenue) was described as a "Shanghai-based digital reading and IP development giant" (e.g., China Literature). This implies MiniMax powers AI narration for audiobooks, podcasts, and content platforms, maintaining a 70% gross margin in its open-platform business.

**Zhipu AI: B-End Focused, Specializing in "Local Deployment"** Zhipu AI follows a classic enterprise service path, positioning itself as a technology enabler. As the first Chinese company to release a 100-billion-parameter model (GLM-130B), it built its business around MaaS (Model-as-a-Service). By June 30, 2025, local deployment (private cloud) revenue hit ¥162 million, making up 84.8% of total income. Its clientele includes state-owned enterprises and large institutions with strict data security needs, serving over 8,000 organizations across tech, finance, and public sectors. GLM model downloads exceeded 45 million in open-source communities.

**Technical Foundations: Multimodal vs. Pretraining Framework** **MiniMax**: Emphasizes multimodal capabilities, covering text (MiniMax M-series), speech (Speech-02), video (Hailuo-02), and music. Its video-generation model Hailuo-02 ranks among the world’s best. Proprietary MoE architecture and linear attention mechanisms reduce inference costs, critical for C-end scalability. **Zhipu AI**: Rooted in academic research from Tsinghua’s KEG lab, it focuses on the GLM-4.5 base model and AutoGLM agent, which simulates human device operations. Its prospectus highlights "safety" and "control," aligning with its enterprise and government clientele.

**Financial Breakdown: Gross Margin as the Litmus Test** **MiniMax**: Gross margin rebounded from -24.7% in 2023 to 23.3% in 2025 (first nine months). Its B-end open-platform business achieved a stellar 69.4% margin, reflecting cost efficiency in model inference and API pricing. Revenue grew from $34.6 million in 2023 to $53.44 million in 2025 (Jan-Sept). **Zhipu AI**: Maintained a 50% gross margin, driven by 59.1% margins in local deployment. However, its MaaS cloud service margin plummeted to -0.4% in 2025 due to price wars.

**R&D "Arms Race": Who’s Burning Cash?** **Zhipu AI**: Spent ¥1.595 billion on R&D in H1 2025 (835.4% of revenue), mostly on computing and talent, signaling aggressive SOTA model pursuit. **MiniMax**: R&D intensity dropped from 2000%+ in 2023 to 337.4% in 2025 (Jan-Sept) as revenue scaled, showing operational leverage.

**Globalization vs. Domestic Focus** **MiniMax**: 26.9% revenue from mainland China, 20.4% from the U.S., with Talkie targeting overseas markets. **Zhipu AI**: Primarily serves domestic clients, with 11.1% revenue from Southeast Asia.

**Capital Backing: Star-Studded Investor Lineup** **MiniMax**: Founders hold control, with Alibaba (13.66%), Tencent (2.58%), and miHoYo (>5%) as key shareholders. **Zhipu AI**: Backed by Ant Group, Tencent, Meituan, Xiaomi, sovereign funds, and VCs like Hillhouse.

**Outlook** With $1 billion (MiniMax) and ¥2.55 billion (Zhipu AI) in cash reserves, both are well-funded for the long haul. Their IPOs present two narratives: - **MiniMax**: Can its super-app model sustain growth and offset compute costs? - **Zhipu AI**: How defensible is its B-end moat against massive R&D spends?

As China’s AI sector moves beyond the "100-model war," these filings mark its entry into commercialization’s deep waters.