By the time Alibaba's Hong Kong-listed shares closed at HK$121.90 on August 13, they had climbed about 36 percent from a June 26 low near HK$89.50, a rebound that erased several months of losses. The explanation most often offered is the one that has been attached to the company all year: AI. Alibaba's artificial-intelligence business is real, fast-growing, and increasingly central to its pitch to investors, and no serious account of the stock can leave it out. But the rally itself is best understood as a change of category, not a change of company. The market has re-bucketed Alibaba from an e-commerce company with a cloud sideline into an AI company with a commerce sideline, and most of the move is the difference between those two buckets. The trigger was a label threshold, and the substance underneath has not yet caught up with it. That gap is the whole story, because it tells you what the stock is priced on right now: not earnings, which are weak, but a label, which is strong.
The threshold that moved the stock
In its March-quarter results, released May 13, Alibaba reported that revenue from AI-related products reached RMB 8.97 billion, roughly 30 percent of its external cloud revenue for the first time, on top of eleven consecutive quarters of triple-digit growth. Management announced two further milestones from the podium: AI revenue would exceed half of cloud revenue within a year, and annualized revenue from AI models and applications would pass RMB 30 billion by the end of the calendar year. The market's response was immediate; the U.S.-listed shares rose sharply during the call after a weak pre-market reaction, and the momentum carried into the summer rally.
None of those numbers is small, and none of them changed the company's profit picture in a single quarter. What changed was the ratio: AI crossed 30 percent of cloud revenue, a milestone that lets an investor say Alibaba is an AI company with a straight face. Thirty percent is a label event, a number chosen by the market's narrative machinery, and the forecast of 50 percent within a year is another, a promise that the label is becoming the substance. The company manages that machinery skillfully, publishing thresholds and meeting them, which is what a company does when the label, not the profit line, is doing the valuation work.
The substance under the label
The rest of the quarter is easy to miss in the excitement, and it matters. Group revenue rose 3 percent, 11 percent like-for-like, mostly outside the core. Adjusted profit fell steeply, non-GAAP net income was barely above zero, and free cash flow was a net outflow for the quarter; for the full year it swung from a large inflow to a substantial outflow, consumed by quick commerce, the consumer AI app, and cloud infrastructure. Commerce still generates the large majority of revenue, and its growth is modest: customer management revenue rose 1 percent against a soft Chinese consumer backdrop.
The same release disclosed the scale of the bet. Analysts have put Alibaba's AI capital program at roughly $56 billion over three years, and the results show what that looks like on the income statement: profit down, cash flow negative, margins compressed. The pressure on the investment case comes from the fact that all of this is happening inside a company whose traditional engine, e-commerce in a weak consumption environment, is barely growing. Every bull and every bear agrees on these facts. They disagree on the bucket.
Two scenarios, one business
The cleanest demonstration that the debate is about categories comes from the sum-of-the-parts analyses published after the results. One widely cited version values the company under four scenarios for the cloud business: as a traditional cloud, at about HK$120 a share; under a conservative AI scenario, at about HK$126; under a base AI scenario, at about HK$138; and under an optimistic AI scenario, at about HK$153. The current price, HK$121.90, sits between the first two, which is to say the market has paid for part, but not all, of the AI story. The same business, the same quarter, the same cash balance, and the entire range of plausible prices comes from the assumption about what the cloud becomes: an infrastructure business growing at single digits or an AI platform compounding at a fifth of revenue growth.
That is the label doing the work earnings used to do. A company used to be worth what its earnings justified; here the multiple is a comment on the category, and the difference between categories is worth roughly 20 percent of the share price. Note what happens inside the generous scenarios: the added value comes from revenue growth assumptions, not margins, because the scenarios assume the company will grow its way to profitability rather than earn its way there. Investors are capitalizing a young flow at a growth multiple, fine while it compounds, uncomfortable the first time it does not.
The benchmark changes with the label
The re-bucketing also changes which benchmarks apply, and that is a quiet form of generosity to the company. Alibaba Cloud's adjusted margin is around 9 percent, and analysts who track the global cloud race note how far that sits below the large Western providers, whose cloud margins run at multiples of it. Measured against global cloud peers, the margin is thin and the comparison uncomfortable. Measured as an AI growth story, it is an afterthought: the growth is triple digits, and the profitability model is supposed to arrive later, after the infrastructure wave matures.
The label decides which yardstick applies, and the choice is worth a fifth of the price. The logic runs in reverse too: value Alibaba as an e-commerce company that spends heavily on AI, and the spending looks like a drag on a slowly growing business; the margin question returns, the multiple falls, and the sum of the parts collapses to the traditional-cloud scenario, roughly where the stock sat two months ago. Both readings are honest readings of the same filings. The market has chosen one, and it can choose the other back; that reversibility is the defining feature of a label-driven price.
What would make the label true
The honest way to hold the stock is to ask what has to happen for the AI bucket to be deserved, and the milestones are not mysterious. AI revenue has to keep compounding toward the majority of cloud revenue, on schedule, without the spending destroying the rest of the business. Cloud margins have to climb from single digits toward something resembling global norms, which in Alibaba's case means a mix shifting toward model platforms, databases, and enterprise software rather than pure compute. Free cash flow, now negative, has to turn positive as the investment wave matures, because a company that burns cash while its legacy engine stalls has a limit, whatever its label. Bulls argue each milestone is plausible, pointing to the dominant share of China's public cloud, the open-source model franchise, its own chip production, and a policy environment that treats domestic AI as a national priority. Bears argue that none of it is assured: the consumer economy is soft, the capex cycle can outlast the demand, and the margin gap versus global peers reflects structural costs of doing AI in China that will not close quickly.
Both cases are arguments about the future, which is why the label matters today. With earnings flat and cash flow negative, the price is being set by the story, and the story is being managed by the thresholds. The interesting part is that the company controls the narrative levers, the reporting of AI revenue, the cadence of milestones, in a way it cannot control the profit line. Labels, unlike earnings, can be staged. That is not an accusation; it is a description of what a re-rating story looks like in practice.
The re-bucketing, and the re-bucketing back
There is nothing unusual about what happened. Markets categorize companies because categories are how prices get set, and a company whose growth engine changes should be re-bucketed. The rally is that work: the market has priced AI revenue growth as the thing that now determines the multiple, and it will keep doing so until the numbers say otherwise. The risk is symmetric, and every category change carries it. Labels get revoked: if the AI thresholds slip, if margins stall, if the consumer economy pulls the commerce majority down hard enough, the market re-buckets, and a re-bucketing is a step function, not a slope. The fall from the AI bucket to the traditional-cloud bucket would not be a slow bleed of multiple points; it would be a return to the old price, because that is what the old category is worth.
For now the label is earning its keep, and the company manages it with the skill of an operator who knows what the market is buying. The question is whether the substance arrives on schedule, quarter by quarter, until the AI majority, the margin climb, and the cash-flow turn make the label a description instead of a hope. Alibaba is not the first company whose price was set by its story, and it will not be the last whose story was judged by its numbers. The story is good; the numbers are next.
Primary sources
- Alibaba Group's March-quarter and fiscal-year 2026 results release, as carried by Business Wire via Nasdaq, for the reported figures on revenue, adjusted profit, free cash flow, cloud revenue growth, AI-related product revenue, the 30 percent milestone, and management's forecasts.
- Sina Finance's August 2026 analysis for the stock's rebound from its June low and for the sum-of-the-parts valuation scenarios spanning the traditional-cloud and AI cases.
- DBS Group Research's May 2026 commentary for cloud margin levels and the comparison with global peers, and Benchmark's reiteration of its rating and target, as reported by Investing.com, for the bull case on cloud growth and profitability.
- Related market reporting for the scale of the AI capital program and the debate over Alibaba's re-rating from e-commerce to AI.