AI Agents Could Unlock Profit for China's Open-Source Models

Chinese artificial intelligence firms have long relied on open-source model releases to accelerate adoption, but thin margins are now driving a strategic pivot toward monetization through AI agents and model-as-a-service (MaaS) frameworks, according to a South China Morning Post analysis.

Alibaba’s Qwen model family, with nearly 1 billion cumulative downloads over three years, exemplifies the success of this open-source strategy. However, as noted in the SCMP report, Alibaba Group Chairman Joe Tsai acknowledged during a November 2025 university talk that the company does not directly profit from its AI models: "We don’t make money from AI, that’s the answer." Instead, Alibaba monetizes through cloud infrastructure and inference services, aligning with a broader industry shift.

The core challenge lies in the economics of open-source models: while they lower barriers to entry, they lack inherent monetization pathways. As SCMP explains, Chinese firms are now embracing hybrid models where open-source foundations coexist with paid services. This mirrors the "commoditising your complement" strategy pioneered by Google with Android, where free software drives adoption of paid ecosystem services like search and advertising.

Alibaba Cloud’s dominance—holding a 36% market share in China’s cloud sector per Omdia research—highlights the viability of this approach. The report notes cloud revenue has become Alibaba’s "biggest growth driver since 2024," following its open-sourcing of Qwen. Similarly, Tencent and other Chinese tech giants are embedding inference services into their cloud offerings, charging enterprises for GPU access, security features, and optimization—effectively monetizing the "stack" around open models.

Analyst Daniel Yue of Georgia Tech’s Scheller College of Business, cited in a LinkedIn analysis, emphasizes that open-source and profit are not mutually exclusive. "The key," he states, "is recognizing AI models are only one layer in a product’s broader stack." This perspective underpins the industry’s move toward AI agents as the monetization engine: agents require continuous inference, creating recurring revenue streams via token-based billing.

Kevin Xu of Interconnected Capital, a former GitHub executive, adds that most users "don’t want to deal with installing, updating, and debugging" open-source models. Instead, they prefer renting GPU-powered inference services—a model Alibaba’s Tsai likened to "staying at a hotel" where customers pay for premium services rather than the core infrastructure.

The shift reflects a maturation of China’s AI ecosystem. Early adopters prioritized speed and scale, but as competition intensifies, firms are refining strategies to balance open collaboration with profitability. The SCMP report notes that while companies haven’t abandoned open-source, "executives emphasized a focus on MaaS during earnings season," signaling a structural change rather than a reversal.

For developers, this means open-source models remain accessible but increasingly serve as entry points to paid ecosystems. Enterprises gain simplified deployment without infrastructure overhead, while firms like Alibaba capture value through scalable cloud infrastructure. As the industry evolves, the distinction between "open" and "closed" models may blur further, with monetization embedded in the service layer rather than the model itself.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn