AI Agent Systems Enter Large-Scale Implementation Phase, Leading Internet Companies Are Significantly Undervalued
AI summary card
AI Agent Systems Enter Large-Scale Implementation Phase, Leading Internet Companies Are Significantly Undervalued
The report indicates that although the China Internet sector faces short-term pressure (YTD -13.9%), AI is evolving from chatbots to multi-agent collaborative systems. Giants like Tencent, Alibaba, and Baidu with full-stack capabilities will dominate value capture, and current valuations (especially after excluding cash) are highly attractive.
- AI evolution has entered the 'agent system' stage: WeChat is about to launch AI agents, which could trigger capital flowing back from hardware to application layers
- Leading companies build high barriers: Compute constraints + chip/cloud infrastructure thresholds make Alibaba, Baidu, and Tencent's full-stack AI capabilities a moat
- Valuation is extremely attractive: Alibaba (ex-cash PE 9.6x), Baidu (5.5x), Pinduoduo (3.3x) are all at historical lows
- Strong cash flows support transformation: The industry as a whole holds substantial net cash, providing solid financial foundation for long-term AI investment
- Preferred Stocks: Tencent (Agent Systems), Alibaba/Baidu (Full-stack Capabilities), NetEase/China Literature/WeGame (AI Games/Content), Manbang/YY (High Profit Growth Certainty)
Report interpretation
Overview
This report systematically reviews the performance of China’s internet industry in Q1 2026 and key trends in the second half of the year. The core conclusion is that the industry is at a critical turning point in the paradigm shift of AI technology — transitioning from single-point models and chatbots to autonomous execution of complex tasks and coordination among multiple agents through 'agent systems'. Although the sector has declined by 13.9% year-to-date and underperformed markets such as Japan (+1.4%), South Korea (-2.4%), and the U.S. (-4.7%), this adjustment is more due to short-term capital flows toward global AI hardware rather than deteriorating fundamentals. The report believes that with improved domestic GPU supply and integration of AI agents into super apps like WeChat, capital will gradually return to the application and connection layer, where top-tier companies with full-stack AI capabilities (chips, cloud, models, applications) and robust cash flows will become the biggest winners.
Core views
The core views of the report revolve around the new stage of AI evolution. First, AI has transcended model competitions to enter a systemized implementation phase centered on 'agents', whose value lies in managing massive agents and automating business processes. If WeChat successfully integrates AI agents, it would serve as a major catalyst attracting capital back to the application ecosystem. Second, full-stack capabilities form an insurmountable moat. Alibaba (including Pingtouge chips, Alibaba Cloud, Tongyi Large Model), Baidu (Kunlun芯, Wenxin Yiyan, AI Cloud), and Tencent (Hunyuan Large Model, WeChat ecosystem) leverage self-developed cloud infrastructure, continuously iterated models, and vast user scenario data to create closed-loop advantages, making them hard to replicate for newcomers. Third, current valuations are severely undervalued, especially after excluding massive net cash, with Alibaba, Baidu, and Pinduoduo at historic lows, reflecting pessimistic sentiment rather than real fundamentals. Fourth, AI is deeply integrating into core businesses: gaming accelerates 3D asset generation and NPC intelligence; e-commerce enhances personalized recommendations and supply chain optimization; advertising uses AI for efficiency gains and cost reduction. Finally, the industry’s profit model shifts from subsidy-driven to sustainable operations, with unit economics clearly improving in sectors like food delivery and FMCG.
Analysis framework
The report adopts a three-layer progressive analytical framework of 'technology evolution – competitive landscape – financial validation'. Layer one focuses on the technical main thread: tracking the iterative pace of Chinese AI players' models (e.g., Qwen 3.7-Max, HY3 Preview, Ernie 5.1), proving they have transitioned from catching up with global frontiers to building proprietary agent systems. Layer two analyzes the competitive landscape: emphasizing compute scarcity and hardware supply chain bottlenecks, deducing that full-stack capabilities are the key determinant of long-term success, not merely single-link advantages. Layer three performs financial validation using hard indicators such as ex-cash PE, free cash flow (FCF), and shareholder returns (repurchases, dividends) to prove that current stock prices do not reflect companies’ true profitability and financial health. The entire analysis anchors technological narratives in quantifiable commercial outcomes, avoiding vague conceptual hype.
Methodology notes
The report treats AI compute power as a core supply constraint, pointing out its scarcity raises industry entry barriers, shaping a winner-takes-all competitive landscape.
In the AI industry, compute power is foundational production material. The report repeatedly emphasizes 'compute power remains constrained,' meaning supply limitations directly prevent new entrants from acquiring sufficient resources to train and deploy large models, naturally protecting giants with self-built clouds and chip capabilities. This is a key premise for understanding their competitive advantages.
The report divides the AI value chain into upstream chips/clouds (hardware), midstream models, and downstream applications/agents, analyzing profit distribution and influence changes across each segment.
The report clearly demarcates the AI industrial chain: upstream includes chips (e.g., Pingtouge, Kunlun芯) and cloud infrastructure, midstream encompasses large models (Qwen, Wenxin, Hunyuan), and downstream involves applications like WeChat and Taobao along with their agents. Its core judgment is that once technologies mature, value concentrates upward (chips/clouds) and downward (user-access defining applications), while pure model players have weak bargaining power.
The report regards free cash flow (FCF) and shareholder returns (repurchases, dividends based on FCF) as core indicators for assessing company intrinsic value and financial resilience.
The report repeatedly stresses 'cash flow and core profits will be key,' listing detailed FCF and shareholder return figures for each company in 1Q26 (e.g., Tencent FCF 8.2 billion, YY shareholder return rate 277%). This shows that under heavy AI investment, whether a company can sustainably generate ample free cash flow is the key differentiator between genuine growth and false concepts.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Tencent Holdings (0700.HK)Top pick, core logic being its WeChat ecosystem as the best carrier for AI agents, with WeChat Agent launch being a major catalyst
- Strengths
- One of the world's largest social platforms with strong user stickiness; rapid iteration of the Hunyuan large model; robust cash flows in gaming and advertising businesses, supporting long-term AI investment
- Weaknesses
- Cloud business constrained by internal GPU priority strategy, slower external commercialization compared to Alibaba and Baidu
- Comparison
- More focused on application-level innovation and user access compared to Alibaba and Baidu, less on underlying chips and cloud infrastructure
- Risks
- Delayed WeChat Agent launch; gaming business growth slowdown; regulatory policy uncertainty
- Alibaba (9988.HK)Top pick, core logic lies in its full-stack capability spanning chips (Pingtouge), cloud (Alibaba Cloud), models (Tongyi), and applications (Taobao, Tmall)
- Strengths
- AI Cloud revenue grows at triple-digit rates for 11 consecutive quarters; leading iterations of Tongyi large model; stable e-commerce base provides cash flow support for AI investment
- Weaknesses
- Intense domestic e-commerce competition pressures margins; international expansion faces geopolitical risks
- Comparison
- Comparable to Baidu in full-stack capability completeness but places more emphasis on AI implementation in e-commerce scenarios (such as AI shopping assistants, AI customer service)
- Risks
- Delayed AI commercialization progress; e-commerce GMV growth stagnation; ongoing antitrust regulation
- Baidu (9888.HK)Top pick, core logic lies in its AI Cloud and large model commercial breakthrough, with a fundamental shift in revenue structure
- Strengths
- AI Cloud revenue now accounts for 52%, far exceeding search ads; self-developed Kunlun芯 chip ensures compute security; Wenxin large model deeply deployed in vertical areas such as search and autonomous driving
- Weaknesses
- Traditional search ad business continues to decline; AI Cloud customers are enterprise-level, with unclear C-end monetization paths
- Comparison
- Leading in AI Cloud commercialization speed over Alibaba and Tencent but lacks ecosystem breadth and user scale
- Risks
- Slowing AI Cloud growth; accelerated decline in search business; delayed mass production and yield of Kunlun芯 chips
Key data
- Sector YTD Change-13.9%Significantly underperforming major markets including Japan (+1.4%), South Korea (-2.4%), and the U.S. (-4.7%)
- Alibaba Ex-Cash PE (FY2028)9.6xSignificantly below historical average, reflecting overly pessimistic expectations
- Baidu AI Cloud Revenue YoY Growth (1Q26)+79%Enterprise demand shifts from model training to large-scale inference deployment, accelerating commercialization
- Baidu AI Revenue Share (1Q26)52%First time exceeding traditional search ads, signaling fundamental business restructuring
- Tencent Cloud GPU Allocation StrategyPrioritize internal needsReflects its AI strategy focusing on empowering its own products (WeChat, advertising, games), rather than purely external cloud services
Impact & implications
The report believes that the rise of AI agent systems will reshape the value focus within the internet industry. Valuation logic previously centered on traffic and user scale will shift toward dual considerations of technical depth (full-stack capabilities) and financial quality (cash flow, net cash). For investors, this means transitioning from 'platform company' thinking to 'technology company' thinking, focusing on those that deeply embed AI into core operations and consistently generate strong free cash flows. In the short term, the sector may continue being affected by macro sentiment and capital rotation; in the medium to long term, with WeChat Agent launch, domestic GPU volume increase, and growing AI application monetization cases, the industry will see significant valuation revaluation and earnings realization double impact.
Risks
- Weak recovery in China's macro economy suppressing consumption and advertising budgets
- Intensified AI token price wars causing industry profitability pressure and valuation downgrades
- Stricter regulatory policies, particularly in AI content generation and data security
- Internet companies used as 'funding sources', facing continuous capital outflow pressure
- Escalating geopolitical risks affecting overseas business expansion and tech collaboration
What to watch
- Official launch date and user feedback of WeChat AI Agent
- Actual delivery and capacity ramp-up progress of domestic GPUs (e.g., Ascend, Kunlun芯) in H226
- Changes in gross margins of Alibaba Cloud and Baidu AI Cloud to validate MaaS (Model-as-a-Service) commercial effectiveness
- Actual monetization progress of Tencent, NetEase, and others in AI gaming (e.g., virtual item sales)
- Free cash flow and shareholder return execution in mid-year reports of internet companies in 2026