Morgan Stanley Maps China's Internet and AI Path: Models, Compute, Cloud, and Applications Drive Industry Re-rating Together
AI summary card
Morgan Stanley Maps China's Internet and AI Path: Models, Compute, Cloud, and Applications Drive Industry Re-rating Together
The report argues that China's AI is developing along the path of improved model capabilities, expanded infrastructure, rising cloud penetration, and deployment of consumer- and enterprise-facing applications, with internet leaders and vertical AI application companies as the main beneficiaries.
- China accounts for an important share of the global top 10 SOTA models, and the report emphasizes that China's model strategy is more open, while mainstream overseas models are more closed-source.
- AI scaling laws are driving compute demand to grow exponentially, and inference, post-training, and long-thought reasoning will materially increase computing resource consumption.
- China's AI competitive advantages come from talent, data, power, policy support, and the domestic internet ecosystem, but U.S. export controls and high-end GPU supply remain the core constraints.
- The application layer spans the WeChat ecosystem, super apps, enterprise GenAI projects, ad content generation, document summarization and rewriting, healthcare, and industrial scenarios.
- The report focuses on AI-related assets and business positioning at Tencent, Alibaba, Baidu, ByteDance, Kuaishou, Meitu, Beisen, and Kingsoft Office.
Report interpretation
Overview
This report is Morgan Stanley's AI-themed investment presentation for China's internet and other services sector, with a core focus on China's AI development path. It covers global model competition, AI infrastructure, China's policy and regulatory environment, public cloud penetration, consumer and enterprise application rollout, and the AI commercialization potential of major internet and software companies.
Core views
The report's core view is that China's AI industry is moving from catching up in model capability toward infrastructure expansion and application commercialization. China has a competitive base in top models, an open model ecosystem, massive user data, developer talent, and policy support; internet platforms have super apps, cloud services, advertising, and content ecosystems that can convert AI capabilities into revenue or efficiency gains more quickly. At the same time, companies such as Baidu may face pressure on traditional advertising from the AI search transition, while high-end chip export restrictions and the progress of domestic GPU substitution remain key risks.
Analysis framework
The report uses a combination of top-down analysis and company case studies: it first compares global SOTA models and the evolution of model performance, then assesses China's endowments, policy environment, AI infrastructure, and public cloud survey data, followed by a breakdown of consumer-facing and enterprise AI use cases, and finally maps the findings to companies such as Tencent, Alibaba, Baidu, ByteDance, Kuaishou, Meitu, Beisen, and Kingsoft Office.
Methodology notes
A layered analysis from foundational models and compute supply to public cloud and then to consumer and enterprise applications.
The report breaks China's AI path into model capability, infrastructure, cloud penetration, consumer applications, and enterprise applications to determine which companies benefit at each stage.
Uses global top model rankings, LMarena Text Score, and the Artificial Analysis Intelligence Index to observe changes in U.S. and Chinese model capabilities.
The report shows China's share among the world's leading models and compares the performance gap and pace of improvement between U.S. and Chinese models.
Measures the impact of AI investment on revenue growth, efficiency gains, and capital returns.
The report separately presents a China AI ROIC calculation and notes that Tencent has strong consumer-facing monetization potential and high-ROI AI deployment opportunities.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Tencentinternet platform and beneficiary of consumer-facing AI monetization
- Strengths
- It has the WeChat ecosystem, super app entry point, and social and content scenarios; the report believes it has the highest consumer-facing monetization potential and relatively high AI ROI.
- Weaknesses
- AI feature commercialization needs to balance user experience, ad load, and regulatory constraints.
- Comparison
- Compared with pure model companies, Tencent's advantage lies in its massive user entry point and multiple revenue conversion paths.
- Risks
- AI application monetization underperforms expectations, insufficient user adoption, and regulatory changes.
- AlibabaAI infrastructure and cloud service enabler
- Strengths
- The report describes it as one of China's best AI enablers, with cloud, models, and enterprise services likely to benefit from higher AI cloud penetration.
- Weaknesses
- Public cloud competition is intense, and there is a time lag between AI capex and revenue conversion.
- Comparison
- Compared with more application-focused platforms, Alibaba is more oriented toward AI infrastructure and enterprise cloud enablement.
- Risks
- Cloud price competition, constrained AI compute supply, and weaker-than-expected enterprise AI budget realization.
- BaiduAI search and foundation model participant
- Strengths
- It has a search entry point and accumulated AI model capabilities.
- Weaknesses
- The report title highlights pressure on its core advertising business from the AI search transition.
- Comparison
- Compared with Tencent and Alibaba, Baidu's main challenge is that its traditional search advertising model is being restructured by AI interactions.
- Risks
- Pressure on search advertising revenue, slow AI search commercialization, and intensifying competition.
- ByteDancefull-stack AI leader
- Strengths
- The report describes it as a full-stack AI leader, covering models and applications such as Volcano Engine, Doubao, Seedance, Seedream, Coze, Jimeng, and Hypic.
- Weaknesses
- Its business coverage is broad, and its AI investment and commercialization path still need ongoing validation.
- Comparison
- Compared with single-use-case companies, ByteDance covers models, cloud, applications, content generation, and hardware entry points at the same time.
- Risks
- Regulation, compute supply, overseas business uncertainty, application retention, and paid conversion.
- Kuaishoumultimodal AI application company
- Strengths
- The report describes it as a leading multimodal player that benefits from video, content generation, and community ecosystems.
- Weaknesses
- Multimodal monetization may depend on the quality of the content ecosystem and advertising budgets.
- Comparison
- Compared with office or enterprise software companies, Kuaishou is more oriented toward content and video generation applications.
- Risks
- Competition among content platforms, homogenization of AI-generated content, and rising compute costs.
- Meituvisual AI application company
- Strengths
- It has AI application deployment scenarios in the visual market and is well suited for monetizing image editing, generation, and creative tools.
- Weaknesses
- The scale of the vertical application market and users' willingness to pay still need ongoing validation.
- Comparison
- Compared with diversified internet platforms, Meitu is more focused on visual creation scenarios.
- Risks
- Increasing competition, slowing user growth, and insufficient subscription conversion.
- BeisenAI application company in the HCM market
- Strengths
- The report positions it as an AI application name in the human capital management market.
- Weaknesses
- Enterprise customers may take longer to adopt AI, and budgets may be slower to materialize.
- Comparison
- Compared with consumer-facing applications, Beisen relies more on enterprise digitalization and ROI in HR scenarios.
- Risks
- Slowing enterprise IT spending, longer project delivery cycles, and intensifying competition.
- Kingsoft Officeoffice software AI application company
- Strengths
- Office scenarios are well suited to document summarization, rewriting, generation, and collaboration, with high-frequency workflow entry points.
- Weaknesses
- AI feature paid conversion and enterprise license upgrades still need to be continuously proven.
- Comparison
- Compared with content entertainment platforms, Kingsoft Office is more oriented toward productivity tools and enterprise software.
- Risks
- Insufficient user willingness to pay, rising model costs, and competition from similar office AI products.
Key data
- China's contribution among the global top 10 SOTA modelsabout halfThe report text states that China contributes half of the global top 10 SOTA models and is the leading competitor outside the United States.
- AI inference compute demand>100X one-shotThe chart indicates that in multi-step long-thought scenarios, inference compute for agentic AI can be significantly higher than a single-pass inference.
- WeChat daily average conversation count44.6 sessions/userThe report cites WeChat user behavior data for July 2025.
- WeChat daily average usage time99.4 mins/userThe report cites WeChat user behavior data for July 2025.
- Typical target-price time frame12-18 monthsThe disclosure notes that Morgan Stanley research typically uses a 12- to 18-month framework for target prices unless otherwise stated.
Impact & implications
For investors, the AI theme affects not only model companies, but also internet platforms, cloud services, advertising, e-commerce content production, office software, human capital management, visual content, short video, and local services. The most direct beneficiaries are platforms with large user funnels and monetization loops, vendors with cloud and model infrastructure, and AI application companies that can demonstrate ROI in vertical use cases.
Risks
- U.S. export controls may limit access to high-end GPUs and advanced compute.
- There is uncertainty around the pace, performance, and ecosystem maturity of domestic GPU substitution.
- AI infrastructure capex may come before revenue realization, depressing short-term returns.
- Public cloud and AI cloud competition may create pricing pressure.
- The AI search transition may disrupt the traditional advertising business model.
- Paid conversion, retention, and ROI for consumer-facing and enterprise AI applications still need validation.
- Data compliance, ethical governance, and changes in regulatory policy may affect product rollout.
What to watch
- China's SOTA models' continued ranking on global leaderboards and the influence of the open-source ecosystem.
- Changes in the share of AI cloud in enterprise public cloud spending.
- CIO survey feedback on GenAI project launch timing, budgets, and ROI.
- Progress in domestic GPU supply, performance, and software ecosystems.
- Usage rates and paid conversion rates of AI features in high-frequency applications such as WeChat, Doubao, Coze, and WPS.
- The contribution of AI investment by internet platforms to revenue from advertising, cloud, content, and enterprise services.