Deep Dive into China's AI Value Chain: Open-Source Model Breakthrough and Computing Power Localization
AI summary card
Deep Dive into China's AI Value Chain: Open-Source Model Breakthrough and Computing Power Localization
Morgan Stanley believes China has captured half of the global AI model market through an 'open-source' strategy and forecasts China’s AI GPU market to reach $67 billion by 2030, highlighting investment opportunities arising from infrastructure advantages that offset technological gaps.
- China contributes more than half of the top 10 globally leading SOTA models, adopting an 'open-source' approach to counter the U.S. 'proprietary' strategy.
- China’s total addressable market (TAM) for AI GPUs is projected to reach $67 billion by 2030.
- Through algorithmic innovations (e.g., MoE, attention mechanism optimization) and infrastructure cluster advantages, China is narrowing its perceived technology gap with the U.S.
- WeChat, with 1.1 billion monthly active users, is viewed as the optimal pre-existing platform for AI Agent development.
- AI-driven revenue already accounts for 43% of Baidu’s total revenue; Alibaba Cloud targets $100 billion in revenue over the next five years.
- Key domestic AI accelerator chip vendors to watch include Cambricon, MetaX, and Iluvatar.
Report interpretation
Overview
This report is part of Morgan Stanley’s 2026 Asia Summer School series, offering an in-depth analysis of China’s end-to-end artificial intelligence (AI) value chain. The core thesis is that despite external constraints, China has secured a significant position in global AI competition through its 'open-source' model strategy, algorithmic efficiency innovations, and robust infrastructure capabilities—contributing over half of the world’s top-tier models. The report forecasts China’s AI GPU market to grow to $67 billion by 2030 and outlines the investment logic and key players across the stack—from foundational chips and cloud infrastructure to 2C/2B applications.
Core views
Model Layer: China counters U.S. 'proprietary' models with 'open-source.' China demonstrates strong competitiveness in state-of-the-art (SOTA) models, contributing more than half of the global top 10. Unlike the U.S.’s closed, proprietary approach, Chinese players such as Alibaba (Qwen) and DeepSeek favor open-sourcing or releasing model weights to rapidly build ecosystems and developer bases. By January 2026, the Qwen model family had surpassed 1 billion cumulative downloads on Hugging Face. Computing Power Layer: Domestic substitution and market growth. China’s total addressable market (TAM) for AI GPUs is expected to grow from approximately $32 billion in 2024 to $67 billion by 2030. Despite restrictions on advanced semiconductor nodes, China is compensating for per-chip performance limitations by scaling mature-node capacity, leveraging chiplet packaging technologies, and deploying large-scale clusters. By 2030, China’s advanced-node capacity is projected to support roughly $58 billion in AI accelerator revenue, with GPU self-sufficiency rising from 41% in 2024 to 86% by 2030. Key domestic vendors include Cambricon, MetaX, and Iluvatar. Infrastructure Layer: Evolution of cloud and data centers. Data center requirements in the AI era differ significantly from the cloud era, with rack power density increasing from 5kW to 10–30kW or higher, and order sizes jumping from MW-scale to 50MW+. China benefits from strong power supply, policy support, and competitiveness in server systems and optical networking. Alibaba Cloud targets $100 billion in cloud revenue over the next five years, implying a CAGR exceeding 40%; Tencent plans to double its new AI investments in 2026 (RMB 18 billion in 2025). Application Layer: WeChat leads in 2C AI Agents; 2B adoption accelerates. WeChat has 1.1 billion monthly active users and an average daily usage time of 99.4 minutes; its ecosystem (payments, mini-programs, official accounts) provides an ideal environment for AI Agent deployment. On the 2B side, industries such as advertising, digital content, finance, and healthcare are beginning early AI adoption, with CIO surveys showing rising AI-related shares in software spending. In Q4 2025, Baidu’s AI-driven revenue exceeded RMB 11 billion, accounting for 43% of its total revenue.
Analysis framework
The report employs a top-down industry chain analysis, combining quantitative forecasting with qualitative comparisons. First, it establishes China’s competitive position in the model layer by comparing U.S. and Chinese model performance, pricing, and strategies (open-source vs. proprietary). Second, it uses a TAM (Total Addressable Market) model to forecast computing demand and evaluates domestic chip self-sufficiency using supply chain capacity data. Third, it analyzes infrastructure investment opportunities by contrasting key metrics between the cloud and AI eras (e.g., power density, delivery timelines). Finally, it leverages user behavior data (e.g., WeChat engagement) and enterprise CIO surveys to map AI application adoption paths and commercial value.
Methodology notes
AI Computing Power Supply-Demand Balance Analysis
The report analyzes the gap between China’s AI GPU demand (via TAM forecasts) and domestic advanced-node supply capacity (e.g., SMIC) to infer the scope for domestic substitution and the trajectory of self-sufficiency improvement—a core framework for understanding semiconductor investment logic.
Ecosystem Network Effects as a Moat
The report emphasizes that WeChat’s massive user base and high-frequency usage scenarios create a natural moat for AI Agent development; this network-effect-based advantage is harder to disrupt than pure technological leadership.
AI Value Chain Transmission Mechanism
The report segments the AI industry into four layers—chips, infrastructure, models, and applications—and analyzes how value flows from底层 hardware to upper-layer software and services, helping investors identify beneficiaries at each stage.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Alibaba (BABA.N/9988.HK)Beneficiary: Owns globally leading open-source model Qwen and robust Alibaba Cloud infrastructure
- Strengths
- Qwen downloads exceed 1 billion; strong cloud revenue guidance; complete ecosystem
- Comparison
- Leads Tencent and Baidu in open-source model ecosystem development
- Tencent (0700.HK)Beneficiary: WeChat ecosystem offers the best landing scenario for AI Agents; increasing AI investment
- Strengths
- 1.1 billion MAU; high social moat; continuous Hunyuan model iteration
- Comparison
- Superior 2C application rollout potential compared to other large tech firms
- Baidu (BIDU.O/9888.HK)Beneficiary: High AI revenue share; leader in autonomous driving commercialization
- Strengths
- 43% AI revenue share; Apollo Go leads in operational mileage
- Comparison
- Leads in AI monetization capability and autonomous driving deployment
- Cambricon (688256.SS)Beneficiary: Core domestic AI chip play; benefits from rising self-sufficiency
- Strengths
- Leading domestic AI chip design capability; policy support
- Comparison
- Alongside Huawei Ascend and Hygon, forms the backbone of domestic computing power
- Runze Technology (300442.SZ)Beneficiary: Surging AI data center demand; high-density rack delivery capability
- Strengths
- Tier 1 market access; liquid cooling and other technical capabilities
- Comparison
- Better adapted to AI-era high-power-density requirements than traditional IDC providers
Key data
- 2030 China AI GPU TAM Forecast$67 billionMorgan Stanley estimate, reflecting long-term market potential
- 2030 China GPU Self-Sufficiency Rate Forecast86%A significant increase from 41% in 2024
- Cumulative Qwen Model Downloads1 billionAs of January 21, 2026, per Hugging Face data
- Baidu AI-Driven Revenue Share43%Q4 2025 data, amounting to over RMB 11 billion
- WeChat Monthly Active Users (MAU)1.1 billionJuly 2025 data, China region
- Alibaba Cloud Five-Year Revenue Guidance$100 billionImplies a CAGR exceeding 40%
- Tencent 2025 AI InvestmentRMB 18 billionPlans to double this investment in 2026
Impact & implications
For internet giants, AI has become a core growth engine; increased capital expenditures by Alibaba Cloud and Tencent Cloud will drive the entire supply chain. For domestic chipmakers, despite technological blockades, substantial local demand and policy support provide a clear growth path, with rising self-sufficiency being a certainty. For application-layer companies, platforms with vast user bases and rich scenarios (e.g., WeChat, Douyin) are best positioned to achieve commercialization of AI Agents. Investors should focus on companies excelling in algorithmic efficiency, infrastructure deployment, and ecosystem integration.
Risks
- Further tightening of U.S. export controls affecting access to advanced chips
- Slower-than-expected AI technological iteration and commercialization
- Intensifying industry competition pressuring margins
- Macroeconomic volatility impacting enterprise IT spending
What to watch
- Quarterly changes in China’s AI GPU self-sufficiency rate
- Actual execution of major tech firms’ AI-related capital expenditures (Capex)
- User penetration and monetization progress of WeChat AI Agent features
- Progress of domestic advanced-node capacity expansion