In November 2019, Disney made a decision that thrilled Wall Street: launching Disney+ to directly challenge Netflix. At the time, Disney was one of the world's most profitable media companies. However, the rise of Netflix created a powerful narrative pressure: cable television was in decline, users were leaving, and if it didn't capture the streaming entry point, Disney's content future would be relegated to being a supplier on others' platforms. Thus, Disney pulled its content from Netflix and invested heavily in its own streaming service. Subscriber numbers soared to over 100 million at one point, but the costs gradually emerged: traditional cable TV business accelerated its decline, ESPN lost subscribers even faster, theatrical release windows were compressed, and the streaming service itself incurred massive annual losses. Disney+ succeeded in user metrics, yet Disney's overall profit margin declined. Bob Iger was brought back, in significant part, to address this situation. The direction of the new business was not wrong. But the profits of the old business were being eroded more quickly by the new venture itself. Sometimes, action is necessary, but effort does not guarantee improvement.
In May 2026, Alibaba faces a structurally similar challenge. On May 13th, Alibaba's Q4 FY2026 results showed that external commercial revenue for Alibaba Cloud accelerated to 40% growth. AI-related product revenue reached 8.971 billion yuan, marking the eleventh consecutive quarter of triple-digit year-over-year growth, with an annualized run rate exceeding 35.8 billion yuan. AI's contribution to external cloud revenue surpassed 30% for the first time. Core e-commerce CMR (Customer Management Revenue) saw comparable growth of 8%. The MaaS platform, Bailian, reported an 8-fold increase in its customer base. T-Head's self-developed GPU chips entered mass production, with over 60% of computing power now serving external clients. In the earnings call, CEO Yongming Wu confirmed: "Alibaba's full-stack AI technology investment has officially moved beyond the initial cultivation phase and entered a positive cycle of scaled commercial returns." These figures are solid. Alibaba has also become the first domestic internet giant to deeply integrate a large language model with its e-commerce ecosystem. On May 11th, Alibaba announced the full integration of its AI model, Qwen, with Taobao. The transition from Qwen's public beta to full platform access took only six months, mobilizing Taobao's entire ecosystem including its 4 billion product database, Cainiao logistics, and flash sales. Pursuing this integration despite a net outflow of 46.6 billion yuan in annual free cash flow demonstrates resolve. However, focusing solely on courage and speed in this narrative risks overlooking critical details.
1. The Three Layers of AI-E-commerce Anxiety What if Taobao hadn't integrated Qwen? In the short term, probably not much. Taobao's monthly active users stand at 951 million, CMR is still growing, and both 88VIP membership and active buyers maintain double-digit growth. The core business is stable. Currently, the impact of AI shopping on the e-commerce landscape remains largely narrative. For instance, Doubao's e-commerce internal testing has just begun scaling, JD.com's AI shopping is in its early stages, and Tencent's Yuanbao is positioned as a decision-aid tool without completing the transaction loop. No AI application has truly captured large-scale GMV from Taobao. Yet, Alibaba's anxiety is real and multi-dimensional. The first layer stems from capital market narratives. Over the past two years, Alibaba has told the market a clear AI story: 380 billion yuan in infrastructure investment (a figure later suggested by Wu Yongming might be an underestimate), Qwen App surpassing 300 million monthly active users, and an 8-fold growth in Bailian customers. This story is already priced into the market. Alibaba's current valuation logic has shifted from "e-commerce profit-driven" to "AI + cloud growth-driven." If the AI narrative lacks evidence of commercialization, the valuation framework could waver. Integrating Qwen with Taobao provides the most tangible landing scenario for the current story: showing the market that AI can genuinely help Taobao sell goods. The second layer of anxiety may come from competitors. Doubao's 345 million MAU solidly leads domestic AI apps; it began internal testing of one-sentence shopping integrated with Douyin E-commerce in late March 2026. JD.com launched its standalone app, JD AI Shop, in late 2025. Tencent's Yuanbao saw daily active users briefly surpass 50 million during the Spring Festival. Kimi has integrated product redirection links for Taobao and JD.com. In this atmosphere, non-participation could easily be misinterpreted by the market as "Alibaba falling behind in AI e-commerce." The third layer may stem from muscle memory. Over the past three years, Pinduoduo and Douyin E-commerce have eroded Taobao's market share from the price and content angles respectively, leading to CMR growth approaching zero or turning negative for multiple consecutive quarters. Alibaba experienced a painful period of deceleration starting in FY2022, with CMR seeing significant negative growth in FY2023, only returning to growth in FY2024. This experience creates a certain muscle memory: it cannot wait for threats to fully materialize before reacting; it must enter the arena early. The urgency behind Qwen's integration with Taobao partly stems from a rational judgment about the AI entry point, and partly from an instinctive reaction of "we cannot afford to be slow again." Layered together, these three anxieties likely drove the rapid and forceful strategic conclusion of "fully integrating Qwen into the e-commerce system." The direction is likely not wrong; the fusion of AI and e-commerce is a definite trend. However, simultaneously ensuring correctness in direction, timing, pace, and implementation form is undoubtedly a high-difficulty maneuver.
2. The Paradox of Efficiency and Dwell Time The structural contradiction touched by Qwen's integration with Taobao will not disappear simply because the direction is correct. The underlying logic of Taobao's business model over the past two decades can be condensed into one sentence: user time is the platform's inventory. Browsing Taobao, comparing prices, watching livestreams, scrolling through recommendations, reading reviews, adding to cart—these behaviors are not necessarily prerequisites for placing an order, but they are Taobao's core weapons. Ad exposure relies on dwell time; recommendation conversion relies on browsing depth; impulse purchases rely on that extra second of attention. Taotian Group's latest quarterly CMR for FY2026 showed 8% comparable growth. Over 100% of its net profit comes from customer management revenue, with other businesses combined still operating at a loss. CMR is Alibaba's cash engine, and one of its engines is precisely "how much time users spend on Taobao." AI shopping pursues the opposite: completing tasks faster, narrowing choices, and placing orders via the shortest path. Qwen's product logic is clear: the user states a request, the AI filters products meeting multiple criteria, completing in thirty seconds a decision that might have taken twenty minutes. A platform that profits from users "wasting" time has now built a tool to help users save time. This contradiction mirrors Disney's dilemma: the better Disney+ performs, the faster ESPN and theatrical audiences decline; the more useful Qwen shopping becomes, the thinner the foundation for Taobao's feed ad exposure becomes. Of course, Alibaba has shown restraint in combining the two. For example, at this stage, Qwen will not prioritize pushing advertised products, indicating Alibaba has chosen a path of "prioritizing experience before monetization." This also means the question of how to earn ad revenue within the AI shopping flow remains unresolved. Traditional e-commerce models are built on a degree of information asymmetry, while AI is inherently a tool to eliminate it. When a user says, "Help me buy the best ergonomic chair," if the AI recommends the best product, the ad system fails; if it recommends the highest-bidding product, user trust collapses. This may represent a structural tension all e-commerce platforms will face between efficiency tools and the attention economy. Alibaba might encounter this problem earlier than others precisely because it is pushing the deepest. The other side of user time is a new test for the advertising system. Taobao's current ad system is the product of two decades of precise evolution. Full-site promotion penetration is still increasing, reaching the expected 30% in FY2026, with a target of nearly 50% for the next fiscal year, clearly a key driver of CMR growth. However, the interactive nature of AI shopping could fundamentally change traffic distribution logic. Once users shop via conversation instead of browsing, "exposure" as a concept may no longer hold. Replacing Taobao's infinite-scroll product feed with three to five AI-recommended products is a difficult transition to smooth out, especially when the old system contributes hundreds of billions in CMR revenue annually. Even with Alimama launching AI Wanxiang to address this change, replacing a profitable old system with a new one that still needs validation carries inherent risks. Recall Disney's lesson: Disney+ burned tens of billions on content, gained over a hundred million subscribers, but the high-margin ad revenue from cable TV eroded even faster. The new business's revenue growth did not outpace the old business's profit decay. The specific problem Taobao faces may not be identical, but the logic shares common ground. The key lies in whether the incremental value brought by AI shopping can cover its erosion of the traditional advertising system.
3. The Mismatch of Core Strengths In terms of interaction characteristics, AI and traditional e-commerce models are fundamentally different. AI shopping excels with standardized products, especially those with few SKUs, transparent pricing, and requiring little emotional decision-making, such as paper towels, charging cables, cat food, and batteries. Spring Festival data corroborates this: 130 million people tried AI shopping for the first time, nearly half the orders came from county-level areas, and nearly 4 million users over 60 used one-sentence ordering. This demonstrates AI's significant value in lowering barriers and reaching new users. However, the standardized product scenario aligns more closely with Pinduoduo's comfort zone than Taobao's. Pinduoduo's product logic aims to reduce decision cost to zero: converged SKUs, focused bestsellers, extreme pricing. Taobao's strength lies in the opposite. Its higher barriers and value reside in categories like apparel, beauty, home goods, trendy collectibles, and interest-based consumption. Shopping behavior in these categories is inherently exploratory, driven by serendipity and desire, not efficiency and the optimal solution. Taobao's heavy investment in content, livestreaming, and short videos in recent years all serve the same purpose: keeping users engaged, creating more opportunities for discovery. "Browsing" is Taobao's moat, but AI is inherently the antithesis of "browsing." Of course, Qwen is already experimenting with AI try-on, product discovery, and outfit recommendations, attempting to prove value in non-standard categories. But there is a significant product evolution gap between an AI that can buy paper towels in thirty seconds and one that can create a sense of "desire" within a conversation. In the matter of AI shopping, Pinduoduo has been playing a certain antagonist role. In their March 2026 earnings call, Pinduoduo management spent considerable time discussing supply chains and rural delivery, but almost none on AI. Their silence raises a question: if AI shopping is truly a promising business, why is the platform most focused on efficiency and low prices in no hurry? ByteDance's Doubao, with 345 million MAU, began internal testing of e-commerce features integrated with Douyin Mall in late March. Doubao can perform cross-platform price comparisons, even stating outright that some products are cheaper on Tmall and JD.com than on Douyin. JD.com launched JD AI Shop in late 2025 with an extremely simple interface retaining only a chat area and recommendation zone. Tencent's Yuanbao integrated with JD.com, WeChat Stores, and Dewu, taking a restrained approach, positioned for decision support without closing the transaction loop. In comparison, Alibaba's moves are indeed the most aggressive. Qwen went from public beta to full platform integration within six months, with a 3 billion yuan Spring Festival promotion driving 130 million users to try it. Among all players, Alibaba resembles the fastest striker on the field. Yet, the scenarios where AI shopping is most likely to succeed in the near term may not align with Taobao's most profitable scenarios, creating the core mismatch in the full integration of Qwen and Taobao. A concern is that Alibaba is expending immense effort promoting AI shopping, potentially validating the direction for the entire industry while reaping limited benefits itself. Following Qwen's integration with Taobao, Alibaba's AI narrative will face new tests. Previously, Qwen's role in boosting e-commerce was primarily an asset of imagination. After full integration, market scrutiny will inevitably sharpen with more precise metrics: AI shopping conversion rates, average order value from the Qwen entry point, AI recommendation return rates, and whether it brings incremental GMV or merely shifts existing volume. These numbers will face greater scrutiny.
4. Entering a Protracted Contest Regarding the relationship between AI and e-commerce, Alibaba's situation is more complex than "lose if we don't act," but also far from "win if we act." Taobao's CMR is growing, users are growing, and the core business is stabilizing. The threat of AI as an e-commerce entry point is real but not yet imminent. The forces driving Qwen's full integration with Taobao likely comprise a mix of strategic judgment about the AI trend, pressure from capital market narratives, urgency created by peer competition, and FOMO (fear of missing out) memories from the last slowdown. These intertwined forces produced a directionally correct but somewhat rushed decision. The internal courage is real. Among all Chinese platform companies, Alibaba is the first to aim its large model directly at its core e-commerce territory, confronting the structural contradiction between efficiency and attention head-on. AI investment is indeed yielding returns: 35.8 billion yuan in annualized AI revenue, 40% external cloud growth, 8-fold Bailian customer growth. AI investment is beginning to translate into operating cash flow. But whether AI shopping itself can become a good business is an unproven proposition. At least for now, the standardized product scenarios it excels in may not be Taobao's most profitable domain; its efficiency-enhancing methods may compress platform dwell time; its inference costs are significantly higher than traditional search; and its compatibility with the existing ad system remains unanswered. The most likely outcome is a middle ground: AI shopping becomes an added capability for e-commerce, not a replacement. It shines in standardized products, slowly penetrates non-standard categories, and coexists with the traditional advertising system long-term rather than rapidly replacing it. For Alibaba, the entry point may be secured, but AI shopping seems unlikely to become a disruptive growth engine in the short term. It resembles a new layer of infrastructure more than a new business model. The business model Taobao has thrived on for two decades is built on users "wasting time." Yet the ultimate goal of AI shopping is likely to help users "save time." How these two objectives coexist within the same entity is Alibaba's core operational challenge for the coming years.
Just as Disney is still seeking balance between streaming and traditional businesses, Alibaba will, for a considerable time, walk a tightrope between efficiency and dwell time, convergence and exploration, proof and restraint. Getting it right could redefine e-commerce. Getting it wrong would be a burdened exploration on behalf of the entire e-commerce industry. The risk in many endeavors lies not solely in being right or wrong, but in doing too hastily, driven by anxiety, something that requires a measured pace.
Comments