A debate over who will emerge as the winner in artificial intelligence is leading to a sell-off of companies with strong fundamentals. As investors shift their focus from steady profit growth to cutting-edge model development and chatbot competition, TENCENT's valuation has fallen to multi-year lows. Since the start of the year, TENCENT's Hong Kong-listed shares have dropped 10%, and are down 21% from last year's peak.

According to analysis, recent research from Bernstein indicates that TENCENT's share price decline is primarily due to a dual blow: perceived lagging progress in AI model development and concerns about AI disruption within the gaming industry. TENCENT currently trades at 14-15 times its projected 2027 earnings, nearing the lows seen in 2022-2023 when the industry faced game approval halts and pandemic lockdowns. Analyst Robin Zhu points out that while TENCENT does need to catch up in chatbot development, the company continues to generate solid returns on its AI investments through its advertising and gaming businesses. The current valuation appears to reflect excessively pessimistic expectations. Meanwhile, Alibaba has gained market favor due to its progress in AI models. With the launch of its Qwen agent service and the release of the Qwen3-Max-Thinking model, Alibaba maintains a leading position in the domestic AI race. Alibaba Cloud's revenue growth of 30-40%, combined with a top-ten global AI model and a leading domestic GPU development project, provides the company with long-term AI optionality. However, analysts question when investors will demand more concrete evidence that agent AI can drive growth in e-commerce GMV share and customer management revenue.

Red envelope campaigns struggle to solve the chatbot dilemma.

Ahead of the Lunar New Year, TENCENT, Alibaba, ByteDance, and Baidu all launched red envelope sharing activities in an attempt to boost user growth for their AI chatbots. Behind this marketing war lies a key issue: for consumers, AI model capability is only one part of the equation. Unlike the 2015 promotion of mobile payments through red envelopes, the logic of using cash incentives to drive user adoption of AI chatbots is more indirect. A decade ago, cash accumulated in users' WeChat wallets directly catalyzed the popularization of online payments and laid the foundation for subsequent transactional ecosystem growth. The closed loop of "giving users money to spend within the same app" makes sense in an e-commerce context—for example, shopping through Tongyi Qianwen. However, it is less clear how red envelope funds drive users to perform information searches or seek AI companionship. QuestMobile data provides some early clues. Alibaba's milk tea subsidies drove Qianwen's daily active users to surge to 73.5 million on February 7, only to fall back to 68 million the next day. TENCENT's Yuanbao app saw its daily active users nearly double to 18.3 million on February 7, before dropping to 16.7 million on February 8. This pattern of sharp spikes and declines raises questions about new user retention. A deeper issue lies in user engagement. While the red envelope campaigns drove significant increases in downloads and daily active users, overall engagement—measured by daily average activated sessions and usage duration—showed limited progress. By the end of January, Tongyi Qianwen had approximately 500,000 high-frequency users (those using it more than 10 times daily), while Yuanbao had about 2 million. These figures remain minuscule compared to major traffic portals like WeChat, Douyin, or Xiaohongshu. Reprogramming human user behavior is a process that takes considerable time.

TENCENT: Underestimated Profitability

Market concerns regarding TENCENT's AI progress are not entirely unfounded. The company's Hunyuan Turbo S model, released in February, ranked only seventh on the LMarena leaderboard, and the company is notably absent from public chatbot rankings. The recent underwhelming performance of the Yuanbao Party event seems to reflect the challenges of a dual-track AI development strategy both inside and outside WeChat. As the messaging infrastructure for 1.4 billion users, WeChat faces a higher quality threshold before introducing new features. However, this narrative pressure appears to be fully reflected in the valuation. Based on Bernstein's forecasts, TENCENT currently trades at 14-15 times its 2027 projected earnings, at the low end of its historical range and near the trough levels seen in 2022-2023 when the industry faced more severe headwinds. More importantly, TENCENT's core operations continue to perform robustly. In the fourth quarter, Sensor Tower data indicated that total mobile gaming revenue declined by approximately 2% year-over-year, but overall gaming growth—including PC games—should remain solid considering the growth from "Delta Action." The advertising business also showed strong performance, with Video Accounts revenue surpassing RMB 10 billion in the fourth quarter, representing roughly 40% year-over-year growth. Bernstein's sensitivity analysis suggests that even assuming a significant increase in operating expenses over the next two years to support AI investments, TENCENT's 2027 earnings per share could still reach RMB 33-35. At that level, the current share price corresponds to a price-to-earnings ratio of less than 15. For context, when Google, Amazon, and Meta have taken turns being labeled "AI laggards," the market has ultimately returned to fundamentals. Robin Zhu suggests that after a period of valuation derating, the psychological model for TENCENT's stock shares some similarities with the US internet sector. The key question is how long investors are willing to wait. Management has hinted that Hunyuan 2.0 is the first model released after a reorganization of the AI development team, and that the pace of model iteration should accelerate thereafter. But until then, the market may continue to focus on the "show me" story of chatbot development.

Alibaba's Big AI Bet

Alibaba's strategy has become clearer in recent weeks. On January 15, Qianwen launched an agent service enabling users to conduct e-commerce transactions, order food, and book travel through the app. The subsequent release of the Qwen3-Max-Thinking model scored around 40 on the Artificial Analysis intelligence index, keeping pace with Minimax and Z.ai and maintaining the company's leading domestic position. The company announced a RMB 3 billion subsidy plan to drive user acquisition and incentivize usage of new features. If all goes according to plan, incremental traffic from Qianwen and the food delivery business should help drive visits to platforms like Taobao, thereby supporting customer management revenue growth. However, the cost is significant. Alibaba reported group capital expenditure of RMB 31.5 billion for the second quarter of fiscal year 2026, while operating cash flow was only RMB 10.1 billion. Bernstein expects the latter to improve in the third quarter and beyond as spending related to food delivery competition moderates. But if investor tolerance for cash burn under the banner of AI capital expenditure diminishes, it could constrain Alibaba management's strategic flexibility. From a valuation perspective, Alibaba's stock has recently traded at a premium to TENCENT's, which Bernstein believes reflects investors applying a sum-of-the-parts valuation to Alibaba Cloud. The analyst's mental framework remains "an e-commerce business worth $100 per share, plus AI optionality." The core e-commerce business generates quarterly NOPAT of approximately $2.0-2.5 per share, providing some downside support for the stock price. The valuation of Alibaba Cloud and broader AI efforts can be debated from multiple angles. However, in a world where Minimax and Z.ai command market capitalizations of $20 billion and Baidu's semiconductor spinoff receives similar attention, the combination of possessing (1) a top-ten global frontier model, (2) Alibaba Cloud with 30-40% revenue growth, and (3) one of China's best GPU development projects clearly holds significant value.