Vertical Integration Enables AI Dominance
MiniMax
The core advantage of Alibaba, Tencent, ByteDance, and Baidu is not just cheaper models, it is that AI can be sold as a feature inside products people already use every day. ByteDance can push Doubao through Douyin and TikTok, Alibaba can attach Qwen to cloud and commerce workflows, and Tencent can place Hunyuan inside WeChat, games, and enterprise software. That lets them buy growth with profits from ads, commerce, and cloud, while MiniMax has to recover model and distribution costs from AI itself.
-
Cross subsidization is concrete. ByteDance generated an estimated $186B of revenue in 2025 from ads, commerce, and live streaming, while also spending heavily on AI infrastructure and pricing Doubao aggressively. A pure play model company does not have a $100B plus cash engine to fund low priced inference and free consumer acquisition.
-
Customer acquisition is built into the product stack. ByteDance has a long pattern of launching new features inside existing apps, promoting them to millions of users, then spinning winners into standalone products. Doubao followed that playbook, reaching nearly 60M monthly users by late 2024, then 157M by August 2025, helped by tight Douyin integration.
-
MiniMax is vertically integrated in a narrower sense. It trains its own text, video, audio, and music models and sells both consumer apps and APIs, which can improve margins and create data feedback loops. But it still lacks the adjacent distribution and infrastructure layers that let platform companies bundle AI into commerce, cloud contracts, social feeds, and gaming traffic.
This pushes Chinese AI competition toward ecosystem battles rather than model only battles. The likely winners will be the companies that combine strong models with owned traffic, cloud capacity, and existing payment relationships. For MiniMax, the path forward is to keep using multimodal product quality and partner distribution to become the specialist layer that platforms and creators still need.