On May 20, 2026, Joe Tsai and Eddie Wu jointly released a letter to shareholders, defining instant retail as the 'primary strategic pillar for upgrading the Taobao and Tmall platforms'.

From a data standpoint, instant retail has indeed demonstrated outstanding performance.

In the fiscal year 2026, Alibaba's instant retail revenue reached RMB 78.52 billion, marking a 47% year-on-year increase and leading the growth charts across all business segments. From January to March, Taobao Flash Sales' order volume surged to 2.7 times that of the same period the previous year, with non-dining retail orders tripling. Hema's Gross Merchandise Volume (GMV) exceeded RMB 107 billion.

The letter contained a notably thought-provoking statement: 'We acknowledge the strategic importance of instant retail in leveraging AI to acquire new users, enhance user engagement, meet diverse consumer needs, and boost transaction volumes and commercialization levels.'

While Alibaba vaguely attributes the business growth of instant retail to 'AI-driven' factors, the actual role of AI remains ambiguous.

Financial reports provide a different perspective. Data indicates that in the fiscal year 2026, Alibaba's sales and marketing expenses skyrocketed to RMB 245.023 billion, a 70.13% year-on-year increase, primarily directed toward customer acquisition for instant retail and user acquisition for the Qianwen App. Meanwhile, free cash flow shifted from a net inflow of RMB 73.87 billion in the fiscal year 2025 to a net outflow of RMB 46.609 billion.

The flow of hundreds of billions in capital has unsettled the market. Financially speaking, the rapid growth of Taobao's instant retail appears to have been fueled by sacrificing group profits and subsidizing millions of 'first cups of milk tea' throughout the year. The precise contribution of AI to this growth remains difficult to quantify.

Shortly before the release of the shareholder letter, Alibaba took another significant step.

Ten days prior, on May 11, Qianwen and Taobao achieved full bidirectional integration. Leveraging Taobao's extensive product library of 4 billion items and over two decades of accumulated real user data, Qianwen enables users to complete the entire shopping journey—from product selection and comparison to checkout and post-sale service—through conversational interactions without redirecting to external links.

The scale of this integration naturally led many to assume that AI would become the core driver for traffic, engagement, and growth on Taobao.

However, the May 20 shareholder letter designated 'instant retail'—a segment already delivering tangible results—as the 'strategic pillar,' while the Qianwen-Taobao integration was briefly mentioned under the 'application layer'.

This contrast adds an intriguing dimension to the Qianwen-Taobao collaboration. The integration seems like a reactive move, akin to 'shooting first and aiming later' in a summary report.

Externally, the growing synergy between Doubao and Douyin E-commerce may have sparked Alibaba's fear of losing its dominant voice in AI-driven e-commerce. Internally, uncertainty persists about the stability of the AI e-commerce model. Consequently, Alibaba has quietly shifted its strategic focus for Taobao toward instant retail, treating the Qianwen-Taobao integration as a contingency plan.

This hesitation is understandable. After all, in the three years leading up to Eddie Wu's aggressive unification of Alibaba's business architecture in 2026, Qianwen and Taobao operated largely independently.

I. The End of a Beginning

The story begins in April 2023.

At that time, Alibaba Cloud launched the Tongyi Qianwen large language model (LLM), with the English name Qwen, under its umbrella. By late October, the Tongyi Qianwen App debuted, led by Zhou Jingren. In May 2024, it was rebranded as the Tongyi App, dropping 'Qianwen' from its name.

However, being part of Alibaba Cloud meant the Tongyi Qianwen App inherently lacked a consumer-centric product mindset. Monthly active users (MAUs) for the Tongyi App languished in the millions.

QuestMobile data reveals that as of September 2025, the Tongyi App had only 3.06 million MAUs, while Doubao dominated with 172 million MAUs—a gap spanning orders of magnitude.

A deeper reason lies in Alibaba's fragmented AI strategy.

Each business unit developed its own AI capabilities independently. Alibaba Cloud gained global recognition for its Qwen series models in open-source communities, while Taobao focused on internal tools like 'Taobao Wenwen' and AI-powered search and recommendation engines to explore new growth avenues. Meanwhile, Quark pushed its AI Super Frame to defend against AI's impact on search engines...

Every business unit duplicated AI efforts to serve its own objectives, neglecting the development of a truly user-centric, brand-building consumer app.

Reports claim that Alibaba's Digital Media and Entertainment Group even paid higher prices for Alibaba Cloud services than external clients did. This organizational siloing not only wasted resources but also delayed Alibaba's AI-to-consumer (AI to C) strategy.

According to reports, during internal discussions before National Day in 2025, Eddie Wu mandated that Alibaba must establish a native AI-powered consumer super app.

Changes came swiftly. On November 13, 2025, media outlets exposed Alibaba's secret launch of the Qianwen project. Four days later, the former 'Tongyi App' was officially renamed 'Qianwen App' and entered public beta testing. The rebranding, decided personally by CEO Eddie Wu, involved over 100 engineers secretly developing the app for months across two dedicated floors at Alibaba's Hangzhou headquarters.

Behind these moves lay Alibaba's undeniable resolve for transformation and its anxiety over dwindling traffic. The company seemed determined to signal to the world: 'This time, we're serious about consumer-facing AI products.'

Resolve, however, is merely a prerequisite for action. Alibaba's weak user acquisition capabilities—a legacy of missing the mobile internet era—continued to haunt it in the AI age. The company chose the most blunt approach to address this flaw: burning cash.

On February 6, 2026, the Qianwen App launched its 'CNY 3 Billion Free Order Extravaganza,' offering free milk tea orders via AI-generated one-click requests.

Within three hours of launch, orders surpassed 1 million; within five hours, they exceeded 5 million. Systems crashed, milk tea shops were overwhelmed, and Hong Kong-listed tea stock prices surged—Guming rose over 5% to a record high, while Chabaidao jumped over 4%. WeChat randomly blocked Qianwen's red packet-sharing links, forcing users to flock to Weibo for deals.

On the first day, Qianwen's daily active users (DAU) skyrocketed from 7.06 million to 58.48 million, a 727.7% surge, 'creating' 51 million daily active users overnight. During the Chinese New Year period, Qianwen's DAU peaked at 73.52 million, with over 130 million participants placing orders nationwide—roughly 1 in every 10 Chinese using Qianwen to shop.

The hype, however, proved short-lived. QuestMobile data shows that by late February 2026, after the campaign ended, Qianwen's DAU dropped to around 30 million, while Doubao maintained a DAU of approximately 100 million—a stark contrast. Analysts noted that with consumer vouchers extended until March 3, Qianwen's true retention rate remained untested.

Latest financial reports reveal that in the fiscal year 2026, Alibaba's 'All Other Segments'—including the consumer-facing Qianwen—reported an adjusted EBITA loss of RMB 35.737 billion, a 276% year-on-year increase. Q4 alone saw a loss of RMB 21.16 billion.

CFO Toby Xu admitted on the earnings call that the negative free cash flow stemmed from heavy AI investments: 'We see this as a historic opportunity with a window of just a few years.'

Qianwen urgently needs a plausible yet compelling monetization narrative to captivate markets.

Taobao, the group's revenue and profit backbone, faces its own challenges.

Previously, Taobao maintained an independent AI strategy. In September 2023, it launched the 'Taobao Wenwen' shopping assistant. By November 2024, Zhang Kaifu, then head of Alibaba International's AI business, returned to Taobao to lead the newly formed 'Search and Recommendation Intelligence Product Division,' overhauling its AI-powered search and recommendation engines.

The B-side efforts were no less significant. In March 2025, Alibaba's advertising LMA model expanded its parameter scale tenfold to reach trillion-parameter levels.

During the 2025 Double 11 festival, Zhang Kaifu revealed that AI had enabled precise understanding and organization of 2 billion products, improving semantic search relevance by 20%, recommendation click-through rates by 10%, and merchant advertising ROI by 12%. That same period, Taobao rolled out six consumer AI products, including 'AI Universal Search' and 'Help Me Choose.'

Zhang Kaifu once remarked, 'Usable AI is the benchmark for Taobao's AI applications.' The subtext: AI is merely a tool; e-commerce remains the goal.

Financially, Q4 fiscal year 2026 reports show that Alibaba's China e-commerce customer management revenue (CMR) grew just 1% year-on-year.

While AI significantly boosted search relevance and recommendation clicks—undeniable product enhancements—systemic pressures like intensified price competition and diminishing traffic dividends prevent these gains from directly translating into financial growth. The improvements, while real, remain insufficient. Unlike Tencent's gaming or ByteDance's advertising, Alibaba's e-commerce core faces multidirectional threats, and its industry leadership is no longer unchallenged.

Taobao's growth hinges on overcoming systemic challenges and industry-wide competition. AI's incremental optimizations alone cannot reverse the tide. Looking back, Taobao's two-year AI deployment now appears to pave the way for Qianwen's commercialization.

After three years of independent operations, these two divisions have finally converged.

II. 'AI E-commerce' or 'E-commerce AI'?

The Qianwen-Taobao integration rests on the ambitious vision of 'conversational commerce.'

Technologically, the closed loop is indeed sophisticated.

Taobao's core capabilities—search, order fulfillment, and after-sales—were encapsulated as standardized skill cards for Qianwen's model to invoke on demand.

When a user requests, 'Find me a summer commute shirt under RMB 200, preferably cotton and light blue,' Qianwen generates styled order cards. Upon user authorization, Alipay is called for payment, and logistics tracking becomes available within the Qianwen interface.

In practice, however, the loop remains unstable. Yet technical glitches are not the most pressing issue—they can be resolved through iterations.

The real question is whether this collaboration can elevate commercial imagination. Commercially, the crux lies in defining the narrative: 'AI E-commerce' or 'E-commerce AI'?

This distinction matters.

'E-commerce AI' treats AI as a tool to enhance e-commerce efficiency. Taobao's AI transformations over the past three years followed this logic, using LLMs to optimize search, ads, and recommendations. While effective, these improvements do not directly translate to revenue.

AI made Taobao more user-friendly, but it did not loosen consumers' wallets.

'AI E-commerce' flips the script. Here, e-commerce is a commercialization scenario for AI, which serves as the gateway and protagonist. Users engage with Qianwen first, generating purchase intent through conversation, with AI completing the transaction. E-commerce becomes one of many real-world applications for AI.

This path reshapes the business model's focus. Alibaba has clearly chosen the latter.

Previously, Taobao's AI-powered search balanced user intent, algorithmic recommendations, and paid bidding—with keyword auctions underpinning CMR.

Now, Qianwen bypasses this system, at least temporarily. This may be a phased product strategy.

If Qianwen's transaction volume scales, commercialization pressures will likely force the reintroduction of advertising auctions. But for now, Alibaba has set aside e-commerce's core monetization method.

This is no small decision. Douyin has yet to integrate Doubao, and WeChat has not embedded Yuanbao as an agent. Among chatbots and super apps, Taobao's move is the most aggressive.

The true challenge lies in the ripple effects of this model shift.

When users no longer manually browse product lists, merchants lose exposure for bid-ranked ads. The shift from organic search traffic to AI-driven recommendations essentially trades a high-margin ad revenue pool for an unproven token-based income model.

Notably, AI may not address the root causes of e-commerce's growth woes.

Alibaba's e-commerce faces systemic challenges: shifting consumer behavior, intensified price competition, and saturated traffic gains. Framing AI-e-commerce integration as a growth panacea risks missing the mark.

Taobao's AI investments appear to be reaching a point of diminishing returns.

In the latest organizational reshuffle, Zhang Kaifu relinquished control of e-commerce AI operations, and the 'Intelligent Search and Recommendation Products' division split into 'Platform User and Product' and 'Intelligent Algorithms' teams. This non-verbal 'grading' of past efforts speaks volumes.

Comparatively, Alibaba is not the only player exploring AI e-commerce, but it is the most determined.

ByteDance's Doubao integrated with Douyin E-commerce in October 2025 and launched its 'Help You Choose' AI shopping feature in May 2026. However, Doubao's shopping journey remains less seamless, and Douyin E-commerce remains content-driven, with users primarily 'browsing' and 'discovering' products.

Tencent's Yuanbao collaborates with JD.com only for AI-guided purchases in select categories like books. Even global giant Amazon has not fully committed, while Pinduoduo remains largely passive on AI.

Unlike its peers, Alibaba possesses both a proprietary LLM and a complete e-commerce ecosystem. Moving forward, the Qianwen-Taobao integration has no blueprint.

Alibaba has rallied under the banner of 'Alibaba dismantles Alibaba,' a romanticized declaration reflecting the giant's do-or-die resolve in this new era.

Amid escalating AI investments and slowing e-commerce growth, Alibaba's path forward seems clear: proactively reshape its core e-commerce gateway and migrate some transaction volume to Qianwen.

Yet, looking back at Alibaba's moves over the past three years since ChatGPT's debut, it is evident that the organizational toll of this arduous journey could have been lighter.

III. Eddie Wu's Resolve to Settle 'Organizational Debts'

On March 28, 2023, the then-Chairman and CEO of Alibaba, Daniel Zhang, unveiled a '1+6+N' organizational restructuring plan in an internal memo to all staff.

Under this new structure, six major business groups within Alibaba Group—Alibaba Cloud Intelligence, Taobao Tmall Commerce, Local Services, Cainiao, Global Digital Commerce, and Digital Media and Entertainment—were set to operate autonomously, establish their own boards of directors, and pursue independent financing and initial public offerings (IPOs) when feasible.

Zhang hailed this move as Alibaba's 'most pivotal organizational reform in 24 years.'

Initially, the markets responded positively, with Alibaba's stock price experiencing an uptick.

However, the tide turned just three months later.

In June 2023, Zhang relinquished his positions as group chairman and CEO, with Joe Tsai taking over as chairman and Eddie Wu stepping in as CEO. By September, Zhang had also stepped down from his role as Alibaba Cloud CEO, effectively bringing an end to the '1+6+N' reform under his leadership.

In November 2023, Alibaba announced that it would no longer pursue the complete spin-off of the Cloud Intelligence Group and that the IPO plans for Hema were on hold.

Objectively speaking, the '1+6+N' spin-off had its external justifications, considering the market environment and expectations at that time. Nevertheless, this spin-off strategy did set the stage for many of the hidden challenges Alibaba now faces in its AI deployment.

The Qianwen App reflects an engineer's perspective, while cultivating user awareness on the consumer side demands a different approach. Alibaba's AI capabilities are predominantly deployed within its own business ecosystem, creating barriers to achieving synergies with Qianwen.

Moreover, although Alibaba's Qwen model is technically on par with, or even surpasses, its competitors and leads the global open-source community, it has yet to present a particularly clear and well-defined collaborative framework to the external world.

The enduring impact of the spin-off on Alibaba's AI development is closely tied to the long-chain nature of large model technology.

The commercial value of large models hinges on a complete 'data-model-user-feedback' loop.

Increased user engagement leads to a smarter product, which in turn attracts more users. This means that the four links of model development, product implementation, user feedback, and commercialization must be seamlessly integrated. Any disruption in any of these links will slow down the entire loop.

When Qianwen functions merely as a model output within Alibaba Cloud, while Taobao & Tmall independently develop their AI capabilities on the other end, user behavior data cannot smoothly feed back into the model training process. This naturally results in reduced efficiency in model iteration and optimization.

Perhaps for this reason, Eddie Wu's core strategy since taking office has been to reintegrate Alibaba. By 2026, the intensity of his unification efforts has nearly reached its zenith.

On March 16, the ATH Business Group was established, integrating Tongyi Lab, the MaaS business line, and the Qianwen Business Unit, with a core goal succinctly captured in nine words: 'Create Tokens, Deliver Tokens, Apply Tokens.'

On April 8, Tongyi Lab was upgraded to the Tongyi Large Model Business Unit, with Zhou Jingren assuming full responsibility. A new Group Technology Committee was established, with Eddie Wu personally leading it, putting an end to the fragmented AI development across different businesses. Within a month, two company-wide communications from Eddie Wu completed the comprehensive centralization of Alibaba's AI efforts.

Centralization, however, does not automatically guarantee synergy. Eddie Wu's directives for integration must be implemented between two massive systems that have operated independently for three years, each with its own technology stack, organizational culture, and interest dynamics.

Today, Qianwen and Taobao are undergoing a forced integration, but unifying technology stacks, aligning data interfaces, and reshaping power dynamics each represent ongoing tests of the organization's collaborative capabilities.

The upcoming 618 shopping festival will serve as the first real battleground for Qianwen's shopping functionality.

If all goes well, conversational shopping will undergo validation for the first time under the traffic pressure of a major promotional event, and Qianwen is expected to truly establish user awareness within Taobao's transaction chain. If the results are unsatisfactory, this integration may devolve into just another 'Taobao Ask,' briefly mentioned in earnings calls before quietly fading away in the next round of organizational adjustments.

Regardless of the outcome, Alibaba has made its heaviest bet yet.

When Qianwen's chat interface supplants Taobao's shopping cart, Alibaba is not merely betting on the success of a single product but on whether the commercial path from 'search advertising' to the 'Token economy' can truly succeed.

The answer will not be fully revealed this 618. Once the direction is chosen, there is no turning back. However, time is precisely what Alibaba does not currently have in abundance.