China’s AI Titans Race for Multi-Dimensional Entry Points
China’s leading technology companies are accelerating their artificial intelligence strategies at an unprecedented pace, launching a wave of new products and platforms that signal a fundamental shift in how the industry competes. From operating system-level AI assistants to open model platforms and general-purpose intelligent agents, the race is on to control what industry experts call the “multi-dimensional entry points” of the AI era.
According to Xinhua News, the competition has moved far beyond surface-level product and traffic battles. It has become what Yan Yang, Vice President of the Beijing IoT Intelligent Technology Application Association, describes as a critical “positioning battle” in the formation of new digital productive forces. “AI entry points are upgrading from the ‘application entry points’ of the internet era to ‘productivity entry points’ in the digital economy era,” Yan told Xinhua.
Four Titans, Four Strategies
Each of China’s major tech players is pursuing a distinct approach to capturing the AI entry point, reflecting different corporate strengths and strategic visions.
Tencent: The OS-Level Gambit
Tencent’s newly launched AI assistant, Marvis (马维斯), operates at the operating system layer—a differentiating factor that sets it apart from most competitors. As reported by 21st Century Business Herald, Marvis integrates terminal systems, files, applications, computing power, and cross-device connectivity into a single AI middleware layer. It comes pre-installed with six specialized agents—PM, File, Computer, Browser, App, and Search agents—that work together as a coordinated team.
Users can issue natural language commands to have Marvis understand tasks, break them down into execution steps, and call upon the appropriate agents. For sensitive operations involving privacy, security, or payments, control is returned to the user. The assistant supports Windows, Mac, and Android platforms with multi-device account synchronization, even allowing remote desktop control from a smartphone.
However, Tencent’s AI ambitions come at a cost. On its Q1 2026 earnings call, CEO Ma Huateng (Pony Ma) offered a strikingly candid assessment: “The AI ship we boarded a year ago is leaking. Randomly grabbing territory will lead to failure.” The company disclosed that new AI products incurred approximately 8.8 billion RMB in losses for the quarter, even as overall revenue reached 196.458 billion RMB (up 9% YoY) with net profit of 59.4 billion RMB (up 19% YoY).
Alibaba: The Platform Aggregator
Alibaba Cloud has taken a different approach with its Bailian (百炼) platform, which the company has now fully opened. Rather than building everything in-house, Alibaba has partnered with leading AI companies including Moonshot AI, MiniMax, Zhipu AI, and Stepfun to aggregate premium large language models into a one-stop service offering “one entry point, multiple model options.”
The strategy leverages Alibaba’s vast ecosystem. Its Tongyi Qianwen (通义千问) model has integrated with Taobao Flash Purchase, Fliggy, AutoNavi, Alipay, and other Alibaba consumer services. A recent full integration with Taobao completed what analysts describe as a critical piece of the ecosystem puzzle, positioning AI as the next-generation super interface for commerce.
The results are impressive: as of March 2026, Alibaba Cloud’s Bailian platform saw an 8x year-over-year increase in customer numbers, signaling accelerating enterprise-level AI demand.
Baidu: The Agent Pioneer
Baidu has launched DuMate (百度搭子), a general-purpose intelligent agent that integrates multiple Baidu products into a single entry point. As detailed by GeLongHui, DuMate is capable of collaborative task execution including intelligent search, code development, deep research, data analysis, and application building.
Baidu’s AI business is showing strong monetization. In Q1 2026, Baidu’s AI business revenue reached 13.6 billion RMB, accounting for 52% of its general business revenue—a milestone that suggests its early pivot to AI is paying off. At the Create 2026 conference, CEO Robin Li proposed a new metric for the AI era: DAA (Daily Active Agents), arguing that the number of agents actively delivering results for humans is a better measure of platform health than raw token consumption.
On the infrastructure side, Baidu’s Kunlun Core P800 chips have completed large-scale validation, with multiple 10,000-card clusters delivered since 2025, achieving 97% effective training rate and over 85% linear scalability at scale.
China Mobile: The Infrastructure Play
State-owned China Mobile has entered the fray with MoMA (移动模型服务平台), a model service platform launched at the 2026 Mobile Cloud Conference. As reported by DoNews, MoMA has integrated over 300 mainstream AI models including its proprietary “Jiutian” foundation model, DeepSeek, Qwen, Doubao, Kimi, and GLM. The platform emphasizes cost reduction and security, targeting government and enterprise clients with features like intelligent routing engines and confidential computing.
The New Rules of Competition
Lu Feng, Dean of the Beijing Frontier Future Technology Industry Development Research Institute, identifies three defining characteristics of the current AI entry point competition: “vertical integration, ecosystem closure, and seamless experience.” Companies are pursuing full-chain self-development from chips to models to applications, using super apps or system layers as containers to embed AI capabilities into user workflows.
“In the context of mobile internet traffic peaking, large models have lowered the barrier to application development,” Lu told Xinhua. “Whoever controls the first interface of user-AI interaction will be able to define rules, distribute traffic, and retain data, thereby gaining the initiative in the new round of competition.”
Chen Liangdong, an analyst at Huayuan Securities, frames the stakes in economic terms: “Entry points equal distribution rights, which in itself constitutes quantifiable strategic value. When intelligent agents can be natively embedded into users’ existing high-frequency entry points, their value is no longer limited to technical capability itself, but lies in the redefinition of traffic distribution mechanisms.”
Industry Chain Ripple Effects
The AI entry point competition is expected to create significant ripple effects across the entire AI supply chain. Massive inference demand will drive cloud GPU and edge-side AI chip shipments. Leading companies may increase procurement of AI servers and optical modules. The personalized processing of massive user data will generate demand for data cleaning, annotation, and vector databases.
Yan Yang noted that this deep competition will produce “powerful siphon and feedback effects, driving the accelerated upgrade of the entire industrial chain.”
What’s Next
The race for AI entry points in China represents more than just product competition—it is a strategic battle to define the next era of human-computer interaction and digital productivity. With major players pursuing distinct strategies at the OS, platform, agent, and infrastructure levels, the outcome will shape not only China’s technology landscape but also global AI development trajectories.
Key questions remain: Will the market consolidate around one dominant AI entry point, or will multiple coexist? How will the tension between platform openness (Alibaba, China Mobile) and vertical integration (Tencent, Baidu) resolve? And can Chinese AI companies achieve profitability while investing heavily in infrastructure and model development?
As Robin Li said at Create 2026: “Self-evolution is a systematic change for the AI era. Only those enterprises that dare to break inertia and continuously reshape themselves will have the opportunity to truly transcend cycles and establish new competitive advantages. There are no bystanders in this era—we are all creators.”