Weigh Alibaba anew.

Core Judgment: Judging from capital expenditure, technology stack, and new revenue sources, Alibaba has already become a pure-blooded AI company; but when looking at profit sources, free cash flow, and user entry points, it remains a transforming enterprise fueled by its e-commerce business. Alibaba has completed the shift of its strategic identity, but its financial identity is still halfway through the transition.

When ChangXin Memory Technologies was pushing for an IPO on the Sci-Tech Innovation Board, a very interesting shareholder appeared in its prospectus: Zhejiang Alibaba Cloud Computing Co., Ltd. It ranks sixth, even ahead of GigaDevice Semiconductor.

In June 2025, Alibaba Cloud Computing invested 6.1 billion yuan to participate in ChangXin's capital increase, securing a 3.85% stake; Alibaba Networks, which invested earlier, still holds 1.12%. Both companies are controlled by the Alibaba Group, with a combined shareholding of 4.97%. Calculated based on the prospectus's valuation methodology, which plans to issue 10% new shares, this stake will be diluted to approximately 4.48% after the issuance. This is the equity relationship explicitly documented in the prospectus.

Why did Alibaba spend 6.1 billion yuan to become a major shareholder of a DRAM manufacturer?

The answer is certainly not limited to a purely financial investment. Large models consume not only GPUs, but also memory, network resources, and data center capacity. The larger the model, the more data needs to be transferred during training and inference processes, making memory bandwidth and supply stability increasingly critical. Alibaba already owns T-Head, cloud infrastructure, and the Tongyi large model. Its further expansion upstream into the storage sector at least demonstrates that its capital allocation approach has changed: in the past, Alibaba invested in traffic, users, and consumption scenarios, but now it is building out the full computing power chain.

However, investing in a chip company cannot automatically turn Alibaba into an AI company, just like buying new energy stocks does not make you an automaker. The real question is: Has AI reshaped Alibaba's capital expenditure, can it become a new revenue engine, and has it rewritten its organizational structure and user entry points?

The latest earnings report presents this issue in a rather tangled manner. Alibaba's revenue for the March quarter reached 243.38 billion yuan, with a mere 3% year-on-year growth; after excluding the previously divested Sun Art Retail and Intime Retail, the comparable growth rate actually stands at 11%. Cloud business revenue hit 41.63 billion yuan, up 38% year-on-year; yet the group recorded an operating loss of 848 million yuan, with Non-GAAP net profit plummeting to just 86 million yuan, barely scraping above the break-even line.

This is not a typical case of "the core business failing". On the contrary, e-commerce monetization is recovering, cloud business growth has re-accelerated, and international operations are approaching the break-even point. The problem is that Alibaba is rapidly reinvesting all new revenue and profits generated from its legacy businesses back into AI, cloud infrastructure, customer acquisition for Tongyi, and instant retail. It is increasingly behaving like an AI company, but its income statement temporarily no longer resembles that of a high-performing enterprise. We will have to wait and see what the next earnings report reveals.

This article will follow the thread of ChangXin's investment to explore: Is Alibaba merely equipping its e-commerce business with AI, or has it fully transformed itself into an AI company sustained by e-commerce cash flow? Along the way, we will conduct a re-evaluation of its overall value.

01. Alibaba's Income Statement Spans Three Eras

To determine whether a long-established internet company has evolved into an AI enterprise, at least three key dimensions need to be examined. First, where is the capital being deployed? Second, where does incremental revenue originate? Third, who will own the next-generation product entry points?

In terms of capital expenditure, Alibaba has already completed its transition. The planned minimum investment of 380 billion yuan in AI and cloud infrastructure over the next three years, the 126.06 billion yuan in capital expenditure for the fiscal year 2026, plus strategic industrial deployments through entities like T-Head and ChangXin, are no longer just about adding a few algorithm teams to support e-commerce. Instead, they represent a full-scale reconstruction of underlying production resources.

On the incremental revenue front, the momentum is clearly shifting toward AI. Alibaba Cloud's full-year revenue reached 158.13 billion yuan, a 34% year-on-year increase, with revenue from external customers growing 33%; by the March quarter, the growth rate of external revenue further climbed to 40%. AI-related products now account for approximately 30% of total external cloud revenue, meaning the business segment with the highest valuation elasticity for Alibaba is no longer e-commerce, but cloud services.

However, Alibaba has not yet completed the transition in terms of profit structure and user entry points. Taobao and Tmall still serve as the primary cash-generating engines, user habits for the Tongyi large model remain unestablished, and the instant retail business is still in a period of heavy investment. A more accurate description would be: Alibaba is already an AI company at the strategic and operational levels, but at the financial level, it remains an internet conglomerate that spans three distinct eras of development.

Taobao and Tmall are mature cash-generating machines. For the fiscal year 2026, Alibaba's China e-commerce business recorded revenue of 449.39 billion yuan, of which customer management revenue was 343.87 billion yuan, representing a 5% year-on-year increase; after deducting revenue offsets generated from new business initiatives, the comparable growth rate was 7%. This set of data at least confirms one fact: Taobao and Tmall's monetization capabilities have not collapsed, and the advertising and software service engine is still operating effectively.

The instant retail and consumer-facing Tongyi businesses are still in the investment phase. They are competing for user frequency, entry points, and the next-generation interaction paradigm, but currently require continuous subsidies, computing power resources, and customer acquisition spending. While legacy businesses remain profitable and new businesses are growing rapidly, consolidated profits are becoming increasingly thin.

For the full year, the adjusted EBITA of Alibaba's China e-commerce group dropped from 193.22 billion yuan to 107.51 billion yuan, a reduction of 85.7 billion yuan; the group's free cash flow shifted from an inflow of 73.87 billion yuan in the previous fiscal year to an outflow of 46.61 billion yuan. Simply put, Alibaba has not suddenly lost its profit-generating capabilities. Instead, it has proactively exchanged a large portion of its profits for two strategic tickets: one leading to the AI future, and the other securing its position in high-frequency consumer scenarios.

Whether these two tickets are truly valuable will ultimately determine if Alibaba is being undervalued by the market, or if it only appears cheap on the surface.

02. E-commerce No Longer Exclusively Captures All Profits

The core reason behind the market's historically low valuation of Alibaba is the perceived erosion of its e-commerce moat: Pinduoduo has captured the value-for-money consumer mindset, Douyin E-commerce has siphoned off content-driven traffic, users' time is increasingly fragmented, and Taobao no longer naturally owns the exclusive entry point for consumer transactions. This judgment has not become entirely obsolete.

However, another critical change cannot be overlooked: the monetization efficiency of Alibaba's e-commerce business is recovering. The growth in customer management revenue and rising penetration of full-platform promotion tools indicate that Taobao and Tmall still control the largest pool of merchant marketing budgets across China. While the traffic dividend has vanished, Alibaba's commercialization system has not failed. As long as GMV does not experience a structural decline, Alibaba will remain a high-margin profit-generating machine.

The problem now is that this profit machine must fund the entire group's transformation. E-commerce profits are not only being allocated to support instant retail, but also to enhance user experience, fund technology R&D, and cover customer acquisition costs for the Tongyi large model.

This is why we cannot simply apply a valuation multiple directly to the 193.2 billion yuan EBITA recorded in fiscal year 2025. That figure represents the upper limit of the old profit pool, not the current freely distributable profit. A more prudent and conservative valuation approach would peg the normalized EBITA of Alibaba's core e-commerce business between 130 billion yuan and 160 billion yuan, assigning it an enterprise value multiple of roughly 8.5 to 10 times. Using this calculation, the core China e-commerce segment would be valued at approximately 1.1 trillion yuan to 1.6 trillion yuan.

03. Instant Retail Is Not the Second Food Delivery Business — It Is Primarily Taobao's Defensive Battle

Interpreting Taobao Flash Delivery merely as "Alibaba entering the food delivery market" would significantly underestimate the strategic importance of this initiative for the group. While food delivery itself is a viable business, Alibaba's core objective is to capture high-frequency user engagement.

Traditional e-commerce relies on low-frequency decision-making: users only need to open the app a few times a month to generate meaningful GMV. Instant retail, however, allows Alibaba to reintroduce coffee, medicine, flowers, snacks, and impromptu purchases back into the Taobao ecosystem. The competition is not for the profit of a single food delivery order, but to return Taobao to the home screen that users open every day. For a company that previously lost ground in lower-tier markets and fragmented user time, this is a critical battle to defend its user entry points.

The operational metrics appear relatively strong. For fiscal year 2026, instant retail revenue reached 78.52 billion yuan, up 47% year-on-year; March quarter revenue hit 19.99 billion yuan, representing a 57% year-on-year increase. The company has also emphasized that order volume, average order value, and unit economics are steadily improving. The unresolved question is how much of this revenue growth is driven by subsidized orders, and whether users will remain after subsidies are reduced — the earnings report has not yet provided sufficient clarity on these points.

As a result, in our valuation framework, instant retail should not be valued as a mature platform from the outset. Under a conservative scenario, we assign zero value, which does not mean the business is worthless, but rather that its value is temporarily offset by ongoing losses. Under neutral and optimistic scenarios, it can be assigned an option value ranging from 80 billion yuan to 150 billion yuan, premised on the continuous narrowing of loss margins, and its proven ability to effectively increase Taobao's purchase frequency rather than merely shifting existing orders within the group's ecosystem.

04. Alibaba's Most Valuable New Narrative: Not Just the Model, But the Full Stack

Major domestic tech companies are all touting their AI strategies, but Alibaba is the closest to becoming a "full-stack AI company": at the infrastructure layer, it owns T-Head and comprehensive cloud infrastructure; at the middle layer, it operates the Tongyi large model, the Bailian MaaS platform, and intelligent agent orchestration capabilities; at the application layer, it integrates Taobao, DingTalk, Amap, Alipay, Fliggy, and the dedicated Tongyi app.

The significance of this architecture is that open-source models do not need to be monetized directly. Tongyi can be offered for free and released as open source, because as long as developers train, fine-tune, and run inference on Alibaba Cloud, every token consumed will eventually translate into cloud service revenue. Joseph Tsai articulated this model very clearly: pure-play model companies find it extremely difficult to profit solely from open-source operations, but full-stack enterprises can recoup value at the cloud infrastructure and application layers.

Financial data is already starting to validate this narrative. For fiscal year 2026, Alibaba Cloud's revenue grew 34% year-on-year, while adjusted EBITA increased by 35%; in the March quarter, revenue from AI-related products reached 8.97 billion yuan, marking the 11th consecutive quarter of triple-digit growth. AI-related products now account for approximately 30% of total external cloud revenue, demonstrating that AI is fundamentally reshaping Alibaba Cloud's revenue structure.

Alibaba previously announced plans to invest a minimum of 380 billion yuan over the next three years to build out AI and cloud infrastructure. Its capital expenditure for fiscal year 2026 hit 126.06 billion yuan. This is a massive investment, but it is not without tangible returns: cloud growth has rebounded from low double digits in the past back to over 30%, the number of Bailian platform customers has surged 8-fold year-on-year, and enterprise model invocation and agent demand are beginning to contribute meaningfully to top-line revenue.

T-Head's strategic value is very real, but it is not yet a domestic equivalent of NVIDIA. Official disclosures indicate that over 100,000 Wuzhuang PPUs have been deployed on Alibaba Cloud's public cloud platform, and are being used by more than 30 automakers and autonomous driving companies. The annual report also notes that its self-developed GPUs have achieved large-scale mass production, covering full lifecycle use cases for training, fine-tuning, and inference. This enhances the controllability of computing power supply, reduces certain inference costs, and provides Alibaba Cloud with an additional buffer against potential chip supply constraints.

However, it currently functions more as a dedicated computing power foundation for Alibaba Cloud, rather than an independent, general-purpose hardware ecosystem like NVIDIA that can be sold to third parties at scale. The appropriate valuation for T-Head should not be derived from hypothetical scenarios like "what would it be worth if spun off for an IPO", but rather based on its tangible ability to improve Alibaba Cloud's gross margins, delivery capabilities, and inference cost-effectiveness. The value of these chips will first be realized within Alibaba Cloud, before any standalone capital market narrative can be justified.

In previous discussions about Alibaba's consumer-facing AI initiatives, an unavoidable criticism was "too many products, but no meaningful user traction". This assessment now needs to be updated: as of March 2026, Alibaba disclosed that the cross-platform monthly active users of its consumer-facing Tongyi services have exceeded 295 million. Judging purely by scale, it is clearly no longer a product that nobody uses.

Third-party data provides a more granular breakdown of these 295 million users. According to QuestMobile statistics, the standalone Tongyi app recorded 166 million monthly active users in March 2026, ranking second behind Doubao. After the peak Spring Festival traffic surge subsided, daily active users stabilized at roughly 30 million, with a quarterly average user activity rate of 17.1%. This scale is already substantial, but compared to Doubao's 33.5% average activity rate, Tongyi still behaves more like a new entry point that has been aggressively acquired through promotions, rather than a daily utility that users have become deeply dependent on.

The cost of this user acquisition is very tangible. QuestMobile reports that Tongyi's download volume reached 170 million in February 2026, with an average monthly download volume of approximately 72.18 million in the first quarter; AppGrowing estimates that Tongyi's advertising spending on app promotion in the fourth quarter of 2025 skyrocketed from less than 100 million yuan in prior quarters to 1.541 billion yuan. The 3 billion yuan "free order" promotion during the 2026 Spring Festival represented additional consumer subsidies, which cannot be directly summed with ad spending, but both clearly demonstrate that Alibaba is investing heavily to secure user entry points. The key concern that requires a more cautious outlook is long-term user retention.

Compared to ByteDance and Tencent, Alibaba's relative disadvantage is that it does not own the top-tier high-traffic user entry points; its core advantage lies in its fully closed transaction ecosystem. While Doubao can only engage users in conversation, and Yuanbao relies on WeChat for distribution, the most viable differentiator for Tongyi is its ability to complete end-to-end tasks directly: selecting products, placing orders, booking hotels, arranging deliveries, and finalizing payments. If this closed-loop ecosystem is fully realized, Tongyi will become the critical user entry point that Alibaba has recaptured after years of absence.

Tongyi has already generated meaningful user traction, but its moat is not yet deep enough. The next priority, more important than simply growing monthly active users, is to improve organic retention rates, usage frequency, and the success rate of AI-powered task completion.

05. International Operations Are No Longer Cash-Burning, and Other Assets Deserve Reassessment

The biggest valuation challenge facing Alibaba's international e-commerce business in the past was not slow growth, but the fact that every additional yuan of revenue generated an equivalent amount of incremental losses. This situation has improved significantly. For fiscal year 2026, Alibaba's international digital commerce revenue reached 144.17 billion yuan, up 9% year-on-year; adjusted EBITA losses narrowed from 15.14 billion yuan to 2.05 billion yuan, with the March quarter recording a loss of only 138 million yuan, essentially touching the break-even threshold.

This means that AliExpress, Lazada, Trendyol, and Alibaba.com are no longer endless cash-burning pits. However, we should not assign them an overly generous valuation: the international business still faces headwinds from local fulfillment complexities, foreign exchange volatility, tariff policies, and intense regional competition, and its revenue growth rate has not yet rebounded to levels that justify a high-growth platform multiple. A valuation range of 100 billion yuan to 180 billion yuan, based on 0.7 to 1.2 times revenue, is a prudent approach.

The "other businesses" segment is more complicated. Freshippo, Cain