Morgan Stanley: China's Consumer AI Monetization Potential Nears ¥300 Billion, Tencent and Alibaba Best Positioned
nashnova research
Morgan Stanley projects China's consumer AI addressable revenue at RMB 294 billion by 2030, with roughly 99% from ads and transaction commissions — not subscriptions. This means → the AI monetization battle is not about who charges users, but who embeds AI closest to a completed transaction.
A RMB 294 billion pie — how is it sliced?
Morgan Stanley breaks 2030 revenue into three layers: transaction commissions RMB 281 bn, advertising RMB 10 bn, subscriptions and pay-per-use just RMB 3 bn.
In plain terms = out of every 100 yuan, 96 come from transaction cuts. Ads and memberships together account for less than 4.
The report states explicitly that this is not net-new market growth but a migration of existing internet revenue into AI-assisted workflows. This reflects Morgan Stanley's core thesis on consumer AI — not a bigger pie, just a different way to slice it.
Why does Morgan Stanley call Tencent the "clearest beneficiary"?
The logic centers on WeChat linking social graphs, Mini Programs, and payments into a single task-execution environment — users go from chat to checkout without leaving the app.
WeChat's AI assistant "Xiaowei" is in a staged rollout, already connected to JD, Meituan, and Trip.com Mini Programs. Its dedicated model has roughly 80 billion total parameters, activating about 3 billion per token.
This means → Tencent does not need users to open a standalone AI app. AI lives inside WeChat, the closest surface to a transaction.
What paths are Alibaba and others taking?
Alibaba's Qwen shopping assistant is embedded in Taobao, covering the full loop from inspiration to after-sales. Morgan Stanley endorses the approach but demands return discipline on standalone Qwen app user-acquisition spend.
Meituan's "Xiaotuan" is shifting from answering questions to placing orders, hailing rides, and booking. Online travel platforms connect itinerary suggestions to inventory; BOSS Zhipin embeds AI in recruitment matching.
In plain terms = every platform is testing the same thesis — make AI complete the action, not just provide an answer.
600 million users already engage with AI — but will they pay?
China's generative-AI user base surged from 249 million in December 2024 to 602 million by December 2025. In Morgan Stanley's survey, 80% of respondents use AI at least weekly for personal tasks, versus 54% in the U.S.
Yet the share who have ever paid for AI dropped from 41% to 35%. Only 21% prefer subscriptions; average maximum willingness to pay is about RMB 41 per month.
This means → user scale is not the bottleneck; willingness to pay is the ceiling. This is precisely the data behind Morgan Stanley's view that "free access + transaction commissions" beats subscriptions.
Four major AI apps — which has the stickiest users?
July 2026 MAUs: Doubao 399 mn, Qwen 161 mn, DeepSeek 124 mn, Yuanbao 51 mn. Doubao's DAU/MAU ratio leads at 42%; DeepSeek sits at roughly 25%.
64% of respondents used at least five AI tools in the past month — multi-platform usage is the norm, and no single app has locked in users.
Morgan Stanley's differentiation call: Doubao wins on traffic iteration, DeepSeek on tech mindshare, Qwen on transaction linkage, Yuanbao on capability incubation within Tencent's ecosystem.
What metrics should investors watch?
Morgan Stanley flags risks: AI's "answer mode" could divert traditional search-ad spend, while AI-generated music and video expand competitive content supply and erode platforms' existing content moats.
The report stresses that profit realization depends on whether incremental revenue can cover model, inference, and marketing costs.
Morgan Stanley recommends tracking three metrics: incremental contribution profit per 1,000 AI tasks, repurchase and retention rates, and free cash flow after capex. In plain terms = don't just watch user counts and GMV — watch how much real profit AI actually adds to each platform.
市场有风险，内容仅供研究参考，不构成投资建议。