At a critical juncture where the global artificial intelligence industry is shifting from a parameter arms race to large-scale application deployment, the AI sector in the Hong Kong stock market is undergoing a profound value reassessment. As 2026 commenced, QINIU (02567) has become a focal point for capital markets, driven by the explosive growth of its MaaS (Model-as-a-Service) platform. On January 19, QINIU officially launched its newly upgraded "AI Large Model Square," marking a significant milestone: this tech giant, long entrenched in audio-video cloud services, is transforming from a traditional cloud service provider into a "Chinese MaaS Scheduling Hub" through its comprehensively covered model square, highly compatible API architecture, and forward-looking Agent+MCP services. Notably, this transformation is not merely a simple business addition but a fundamental restructuring of the underlying development paradigm for the AI Native era. As the only scarce target in the Hong Kong stock market possessing "AI hub" capabilities, QINIU is filling the gap in the "model scheduling layer" within China's AI ecosystem. Its benchmark is not just a simple cloud service model but the globally sought-after "OpenRouter" model. This scarcity stems from its absolutely neutral positioning and powerful bilateral network effects, propelling the company's value into a new reassessment cycle.
In 2025, the large model market landscape was in a state of intense and dynamic flux, with the rise of open-source models fundamentally altering the industry's rules. User habits shifted from reliance on a single model to switching, combining, and using models in parallel across different scenarios. QINIU astutely captured this trend, using its MaaS platform to break down the physical barriers between various model vendors and providing developers with a unified scheduling entry point. Importantly, unlike giants like AWS and Azure, which are deeply tied to the interests of specific model vendors, QINIU maintains absolute neutrality, meaning it does not represent the interests of any particular model vendor, thus becoming the ultimate platform for developers to conduct A/B testing and performance benchmarking. Developers can directly configure QINIU Cloud MaaS's API Key in tools like VS Code or Cline to seamlessly switch between models such as DeepSeek or GLM based on business needs, without complex refactoring work. The architectural advantage of this "Chinese version of OpenRouter" allows it to share the红利 of exploding model invocation volume through a "pay-as-you-go" premium model, without bearing the high losses associated with model research and development.
Since the launch of the MaaS platform in 2025, the number of related users has shown non-linear expansion, exceeding 180,000. As of January 14, 2026, the total number of registered users on the QINIU Cloud platform has officially surpassed 1.92 million. In just the most recent months, the company has added over 200,000 new developers and enterprise users, a growth rate signifying that its MaaS ecosystem has already generated significant scale effects. Driven by both policy and technology, the AI赛道 exhibits characteristics of a "long slope and thick snow," and QINIU, with its neutral multi-model comparison capability, has secured a first-mover advantage in the fiercely competitive market.
Looking ahead to 2026, capital markets widely regard it as the breakout year for Multi-Agent reasoning. As production paradigms shift from simple dialogue to long-sequence, multi-step complex reasoning tasks, multi-model scheduling will become an inevitable requirement during agent task execution. Under this trend, the invocation frequency and context consistency requirements for MaaS platforms will grow exponentially. In anticipation, QINIU has proactively laid out its Agent+MCP services. By providing pre-configured, standardized MCP plugins, it empowers large models with the ability to "perceive the world" and "execute tasks," not only breaking the physical boundaries of AI but also making it the preferred foundation for building AI Agents. In the future, as next-generation models like DeepSeek V4 further reduce inference costs, the invocation frequency of AI Agents will experience a true explosion. As the technical foundation supporting millions of users, QINIU's value in model distribution and scheduling will be amplified unprecedentedly.
More importantly, QINIU is building a deep data moat based on its MaaS platform. Through a full-stack management console, enterprises can achieve fine-grained control over resources, intuitively comparing the generation quality, latency, and response speed of different models under the same prompt, thereby saving budget and optimizing costs. Unlike performance data benchmarked by the vendors themselves, QINIU provides a true profile based on developers' actual invocation success rates, time-to-first-token, and response speeds. For enterprise customers, this transparent,精细化 resource control capability holds immense commercial appeal. The data feedback generated by massive real-world invocations forms QINIU's unique "data flywheel": the more users, the more accurate the model evaluations; the more accurate the evaluations, the more it attracts developers to use the best-fitting model rather than the nominally strongest one; the resulting strong user stickiness is a barrier difficult for pure technology service providers to cross.
From a broader industrial perspective, QINIU stands at the convergence point of three overlapping waves: "audio-video + AI + cloud services." In the first half of 2025, the company's AI-related revenue exceeded 184 million yuan, contributing 22.2% of its total revenue, a marginal change confirming the effectiveness of its transition from traditional audio-video cloud services to AI infrastructure. In the first half of 2025, the scale of China's MaaS market reached 1.29 billion yuan, a year-on-year increase of over 400%, indicating the赛道 is in a phase of explosive growth. In this blue ocean, QINIU's scarcity as the only target in the Hong Kong market with AI hub attributes becomes increasingly prominent. Meanwhile, capital market attention to this scarce target continues to heat up, with institutions such as Guotai Haitong and Shenwan Hongyuan previously issuing "Overweight" ratings for QINIU.
As multi-agent applications enter a period of large-scale deployment in 2026, market reliance on the "model scheduling layer" will grow exponentially. As this ecosystem effect takes shape, QINIU's business model is evolving from a simple "transaction fee" model to a strategic升维 of "ecosystem empowerment," and its valuation anchor will shift from traditional Price-to-Sales ratios to a more imaginative platform premium. The current QINIU is gradually evolving from a mere service provider to an ecosystem core, carving out a unique path amidst a landscape dominated by giants. As the second half of AI shifts towards an efficiency game, its core value as an "AI hub" is expected to accelerate its realization, indicating that its value reassessment in the capital market has only just begun.
Comments